Outcomes of colonization with MRSA and VRE among liver transplant candidates and recipients.

Methicillin-resistant Staphylococcus aureus (MRSA) and vancomycin-resistant enterococcus (VRE) infections cause significant morbidity and mortality among liver transplant candidates and recipients. To assess rates of MRSA and VRE colonization, we obtained active surveillance cultures from 706 liver transplant candidates and recipients within 24 h of admission to an 11-bed liver transplant ICU from October 2000 to December 2005. Patients were followed prospectively to determine the cumulative risk of MRSA or VRE infection or death by colonization status. Outcomes were assessed by Kaplan-Meier survival analysis and Cox regression and multivariate logistic regression adjusting for covariates. The prevalence of newly detected MRSA nasal and VRE rectal colonization was 6.7% and 14.6%, respectively. Liver transplant candidates and recipients with MRSA colonization had an increased risk of MRSA infection (adjusted OR = 15.64, 95% CI 6.63-36.89) but not of death (adjusted OR = 1.00, 95% CI 0.43-2.30), whereas those with VRE colonization had an increased risk both of VRE infection (adjusted OR = 3.61, 95% CI 2.01-6.47) and of death (adjusted OR = 2.12, 95% CI 1.27-3.54) compared with noncolonized patients. Prevention and control strategies, including use of active surveillance cultures, should be implemented to reduce the rates of both MRSA and VRE colonization in this high-risk patient population.

Use of quantitative antibiogram analysis to determine the clonality of coagulase-negative Staphylococcus species from blood culture.

Two phenotypic methods, quantitative antibiogram analysis and colony morphology, were compared to pulsed-field gel electrophoresis (PFGE) for distinguishing the clonality of coagulase-negative Staphylococcus (CNS) species. The results of these three methods were correlated with the patients' clinical findings for 23 episodes in which CNS species were isolated from two blood culture bottles within a 24-h period. Quantitative antibiogram and colony morphology at 24 h correlated with PFGE typing in 21 (91%) and 20 (87%) episodes, respectively. All episodes associated with CNS strains with identical PFGE patterns had quantitative antibiogram similarity coefficients < 10, whereas most episodes associated with strains with different PFGE patterns had quantitative antibiogram similarity coefficients >or= 17. The CNS isolate pairs were less likely to be associated with infection if the strains had different PFGE types or a quantitative antibiogram similarity coefficient >or= 17. Clinical microbiology laboratories should consider use of the quantitative antibiogram similarity coefficient to aid clinicians in distinguishing infection-associated CNS blood isolates from contaminants.

Vancomycin-intermediate Staphylococcus aureus in a home health-care patient.

In June 2000, vancomycin-intermediate Staphylococcus aureus (VISA) was isolated from a 27-year-old home health-care patient following a complicated cholecystectomy. Two VISA strains were identified with identical MICs to all antimicrobials tested except oxacillin and with closely related pulsed-field gel electrophoresis types. The patient was treated successfully with antimicrobial therapy, biliary drainage, and reconstruction. Standard precautions in the home health setting appear successful in preventing transmission.

Laboratory-based surveillance for vancomycin-resistant enterococci: utility of screening stool specimens submitted for Clostridium difficile toxin assay.

OBJECTIVE: To study vancomycin-resistant enterococci (VRE) gastrointestinal colonization prevalence in high-risk hospitalized patients and to assess the cost and utility of this laboratory-based surveillance. SETTING: Large university teaching hospital. DESIGN: Quarterly prevalence culture survey of 50 stool specimens submitted for Clostridium difficile toxin A assay from October 1996 through June 1999 (n=526). Screening culture survey of all C difficile-positive stool specimens from July 1998 through June 1999 (n=140). PATIENTS: Specimens for analysis were collected from patients who were admitted to the hospital and who had C difficile toxin A testing ordered. Patient samples were excluded from analysis if they were obtained from patients not hospitalized at UCLA Medical Center, if the C difficile toxin assay result was indeterminate, or if the patient was known to have previous VRE colonization or infection. RESULTS: During quarterly surveillance, VRE was detected in 19.8%, C difficile toxin A in 9.5%, and both VRE and C difficile toxin A in 3.2% of stool specimens submitted for C difficile toxin assay. Patients whose stool specimens were positive for C difficile toxin A were significantly more likely than those whose specimens were negative to have VRE detected (odds ratio, 2.3; 95% confidence interval, 1.2-4.5). Based on these findings, in July 1998, we began routine screening of all C difficile-positive stool specimens for VRE. From July 1998 through June 1999, 58 (41.4%) of 140 patients with C difficile-positive specimens had VRE newly detected in the stool. The combined cost of the two laboratory-based surveillance strategies was approximately $62 per VRE-positive patient identified and $5,800 per year. CONCLUSION: Quarterly surveillance of stool submitted for C difficile assay combined with screening all C difficile-positive stools is a cost-effective and efficient strategy for detecting VRE stool colonization among high-risk hospitalized patients. Such a laboratory-based surveillance should be included as part of a comprehensive program to limit nosocomial VRE transmission.

Outbreak of Stenotrophomonas maltophilia bacteremia in allogenic bone marrow transplant patients: role of severe neutropenia and mucositis.

From March 1997 through November 1997, 8 allogenic bone marrow transplant (BMT) patients developed Stenotrophomonas maltophilia bacteremia on the hematology service at UCLA Medical Center (Los Angeles). Five of these patients had undergone transplantation during the same hospitalization that S. maltophilia bacteremia was detected (case patients). Compared with 7 concurrently hospitalized allogenic BMT patients (control patients), the 5 case patients were more likely to have been hospitalized in room A (P=.045), to have severe neutropenia on the culture date (P=.028), to have a longer duration of severe neutropenia (P=.05), to have severe mucositis (P=. 028), and to have received total parenteral nutrition (P=.028). Pulsed-field gel electrophoresis revealed that 2 of 3 isolates from case patients hospitalized in room A were identical. In allogenic BMT patients, severe neutropenia and severe mucositis may promote infection with S. maltophilia by impairing host defenses.

Are high-velocity tricuspid and pulmonary regurgitation endocarditis risk substrates?

BACKGROUND: A major predisposing cause of infective endocarditis is a susceptible cardiac substrate characterized by high-velocity turbulent flow. However, the risk incurred by high-pressure, high-velocity regurgitation across inherently normal pulmonary and tricuspid valves has not hitherto been examined. METHODS AND RESULTS: This study focused on 186 adult patients with congenital heart disease who had pulmonary vascular disease and inherently normal right-sided pulmonary and tricuspid valves. The observation period was approximately 1646 patient-years. Exclusion criteria were coexisting lesions that might have served as independent risk substrates for infective endocarditis. High-velocity turbulent pulmonary and tricuspid regurgitation were identified and quantified by color flow imaging and continuous wave Doppler echocardiography. Diagnoses of infective endocarditis were based on established clinical and laboratory criteria. Tricuspid regurgitation was moderate to severe in 80 patients and mild or absent in 106 patients. Pulmonary regurgitation was moderate to severe in 84 patients and mild or absent in 102 patients. With the exception of a single habitual intravenous drug abuser, no patient, irrespective of the degree of high-velocity turbulent pulmonary or tricuspid regurgitation, had infective endocarditis. CONCLUSIONS: High-velocity turbulent flow across inherently normal pulmonary and tricuspid valves rendered incompetent by pulmonary hypertension may represent a relatively low-risk or no-risk substrate for infective endocarditis.

Something's rotten: a nosocomial outbreak of malodorous Pseudomonas aeruginosa.

From July 1994 through November 1996, a phenotypically unique strain of Pseudomonas aeruginosa producing a pungent, "rotten-potato" odor and a positive lysine decarboxylase reaction was isolated from 39 patients at UCLA Medical Center (Los Angeles). Most cases (95%) were in intensive care units and had clinical infections (72%). Most isolates (74%) were recovered from cultures of respiratory secretions. To determine risk factors for acquisition of the organism, 23 cases were compared with 23 randomly selected controls matched by service and isolate date. Multivariate analysis revealed that isolation of malodorous P. aeruginosa was associated with mechanical ventilation of > 24 hours' duration (odds ratio [OR] = 9.4; P = .001) and transfer from an outside hospital (OR = 5.7; P = .04). DNA from outbreak strains hybridized to P. aeruginosa-specific toxin A and phospholipase C gene probes and all outbreak isolates tested were found to be identical by use of pulsed-field gel electrophoresis. An unusual phenotypic characteristic of the strain led to the recognition of a nosocomial outbreak of P. aeruginosa infection associated with mechanical ventilation.

The epidemiology of resistance to ofloxacin and oxacillin among clinical coagulase-negative staphylococcal isolates: analysis of risk factors and strain types.

Coagulase-negative staphylococci are important nosocomial pathogens that increasingly are resistant to oxacillin and fluoroquinolones. To determine predictors of acquisition of oxacillin and ofloxacin resistance, we prospectively identified 150 patients from whose clinical specimens coagulase-negative staphylococci were isolated that differed in susceptibility to oxacillin and ofloxacin. In multivariate analysis, isolation of ofloxacin-resistant coagulase-negative staphylococci was associated with receipt of aminoglycosides (odds ratio [OR] = 8.45; 95% confidence interval [CI] = 2.10-34.1; P = .001) and fluoroquinolones (OR = 11.50; 95% CI = 4.15-31.6; P < .001) within 30 days; oxacillin resistance was associated with prior receipt of beta-lactam agents (OR = 5.99; 95% CI = 2.91-12.3; P < .001). Among oxacillin-resistant strains, there was heterogeneity of pulsed-field gel electrophoresis (PFGE) types, and no type was common between ofloxacin-resistant and ofloxacin-susceptible strains. Thus ofloxacin resistance may have emerged de novo among diverse oxacillin-resistant strains following the selection pressures of antimicrobial therapy. In contrast, 50% of patients with oxacillin-susceptible/ofloxacin-resistant strains had one of two PFGE types, a finding suggesting that person-to-person transmission resulted in the dissemination of some of these strains.

Emergence and dissemination of a highly vancomycin-resistant vanA strain of Enterococcus faecium at a large teaching hospital.

We prospectively identified patients at the Massachusetts General Hospital from whom vancomycin-resistant enterococci (VRE) were isolated from a clinical specimen from 1 January 1991 through 31 December 1995. VRE strains were available from 139 (82%) of the 169 patients with clinical cases. Of these, 39 (28%) were identical or closely related by pulsed-field gel electrophoresis (i.e., VRE type A strain), including 38 (43%) of 89 VRE strains in 1995. By multivariate analysis, acquisition of the VRE type A strain was associated with receipt of clindamycin (odds ratio [OR] = 10.5), 15 or more days of hospitalization before the first isolation of VRE (OR = 2.9), and residence on one of the general medical floors (OR = 7.8). The VRE type A strain was a vanA strain of Enterococcus faecium and was highly resistant to all antimicrobial agents tested except chloramphenicol. These findings document the rapid dissemination of a highly resistant strain of E. faecium among patients and among other extant VRE strains at the Massachusetts General Hospital in 1995.

Transcriptional activation of Salmonella typhimurium invasion genes by a member of the phosphorylated response-regulator superfamily.

The Salmonella typhimurium PhoP-repressed locus prgHIJK encodes components of a sec-independent type III secretion apparatus. This apparatus is composed of at least 17 proteins encoded on a 40 kb pathogenicity Island located at centisome 63 on the S. typhimurium chromosome. The secretion apparatus and some of its targets, SapB, SapC and SspD, are necessary for epithelial cell invasion. The transcription of many invasion genes, including prgHIJK, is coordinately activated by HilA, a transcription factor encoded within the pathogenicity island. In this report we identify sirA, a gene located outside the pathogenicity island that is essential for induction of prgHIJK and hilA transcription. sirA encodes a 234-amino-acid protein that is essential for S. typhimurium Ssp (Salmonella secreted protein) secretion and invasion and is similar to response regulators of two-component regulatory systems. sirA-mutant phenotypes could be suppressed by two DNA clones from unlinked loci, designated sirB and sirC. These data suggest that SirA may be phosphorylated in response to S. typhimurium sensing a mammalian microenvironment. Furthermore, SirA phosphorylation is predicted to initiate a cascade of transcription-factor synthesis which results in invasion-gene transcription, Ssp secretion, and bacterial invasion of epithelia.

Implementation and evaluation of an algorithm for isolation of patients with suspected pulmonary tuberculosis.

OBJECTIVE: To implement and evaluate an algorithm designed to assist in the consistent placement of patients with suspected pulmonary tuberculosis into negative-pressure isolation rooms (NPIRs). DESIGN: A standard algorithm was designed for the appropriate room placement of patients with suspected pulmonary tuberculosis using clinical, radiographic, and laboratory criteria and reported risk factors. A case-patient was defined as an inpatient who had at least one Mycobacterium tuberculosis culture-positive respiratory specimen from January 1, 1993, through December 31, 1994. Demographic, clinical, laboratory, case contact, and isolation and room placement data were collected prospectively on all case patients. SETTING: A 900-bed university teaching and referral center. RESULTS: During 1993 and 1994, 69 patients were evaluated for possible pulmonary tuberculosis, and 31 case-patients were identified. Of the 31 case-patients, 26 (84%) were placed on respiratory isolation in NPIRs, including 19 (61%) who were isolated within 24 hours of admission (1993, 14 of 20 [70%]; 1994, 5 of 11 [45%]). Seven case-patients (23%) were isolated in NPIRs following delays that ranged from 2 to 31 days (median, 9 days), and five case-patients (16%) never were isolated during admissions of from 3 to 28 days (median, 4 days). These 12 case-patients contributed a total of 136 patient-exposure days during their hospitalizations. Misclassification of patient risk status by user error delayed isolation of five (42%) of the 12 improperly isolated case-patients. CONCLUSIONS: The use of an algorithm incorporating radiographic, laboratory, and clinical criteria and reported risk factors may assist in the rapid isolation of patients with suspected pulmonary tuberculosis.

Burkholderia cepacia respiratory tract acquisition: epidemiology and molecular characterization of a large nosocomial outbreak.

In 1994 we investigated a large outbreak of Burkholderia (formerly Pseudomonas) cepacia respiratory tract acquisition. A case patient was defined as any patient with at least one sputum culture from which B. cepacia was isolated from 1 January to 31 December 1994. Seventy cases were identified. Most (40 [61%]) occurred between 1 February and 31 March 1994; of these, 35 (86%) were mechanically ventilated patients, 30 of whom were in an intensive-care unit (ICU) when B. cepacia was first isolated. Compared with control patients who were mechanically ventilated in an ICU, these 30 case-patients were significantly more likely to have been ventilated for 2 or more days (30/30 v. 15/30; P < 0.001) or to have been intubated more than once (12/30 v. 2/30; OR = 9.3, 95% CI 1.6-68.8; P = 0.002) before the first isolation of B. cepacia. By multivariate analysis, the 35 mechanically ventilated case-patients were significantly more likely to have received a nebulized medication (OR = 11.9, 95% CI = 1.6-553.1; P < 0.001) and a cephalosporin antimicrobial (OR = 11.9, 95% CI = 1.6-553.1) in the 10 days before the first isolation of B. cepacia, compared with B. cepacia-negative control-patients matched by date and duration of most recent mechanical ventilation. Although B. cepacia was not cultured from medications or the hospital environment, all outbreak strains tested had an identical DNA restriction endonuclease digestion pattern by pulsed-field gel electrophoresis. Review of respiratory therapy procedures revealed opportunities for contamination of nebulizer reservoirs. This investigation suggests that careful adherence to standard procedures for administration of nebulized medications is essential to prevent nosocomial respiratory infections.

PhoP/PhoQ transcriptional repression of Salmonella typhimurium invasion genes: evidence for a role in protein secretion.

Previously, the PhoP-repressed locus prgH was identified as important for signalling epithelial cells to endocytose Salmonella typhimurium. Characterization of prgH revealed that it is an operon of four genes encoding polypeptides of 392 (prgH), 80 (prgI), 101 (prgJ) and 252 amino acid residues (prgK). Synthesis of the 2.6 kb prgHIJK transcript was repressed in bacteria that activate PhoP/PhoQ, indicating that PhoP/PhoQ regulates prgHIJK by transcriptional repression. The prgI, prgJ and prgK predicted gene products were similar to Shigella flexneri and Yersinia enterocolitica proteins required for secretion of Ipa and Yop virulence factors. Analysis of the culture supernatants from wild-type S. typhimurium demonstrated that at least 25 polypeptides larger than 14 kDa could be detected. In contrast, prgH1::TnphoA, phoP-constitutive and hil-deletion mutants had significant defects in their supernatant protein profiles. The invasion and supernatant protein profile defects of the prgH1::TnphoA mutant were both complemented by a 5.1 kb plasmid that included prgHIJK. These results suggest that PhoP/PhoQ regulates extracellular transport of proteins by transcriptional repression of secretion determinants and that secreted proteins may be involved in signalling epithelial cells to endocytose bacteria.

Postoperative infections traced to contamination of an intravenous anesthetic, propofol.

BACKGROUND: Between June 1990 and February 1993, the Centers for Disease Control and Prevention conducted investigations at seven hospitals because of unusual outbreaks of bloodstream infections, surgical-site infections, and acute febrile episodes after surgical procedures. METHODS: We conducted case-control or cohort studies, or both, to identify risk factors. A case patient was defined as any patient who had an organism-specific infection or acute febrile episode after a surgical procedure during the study period in that hospital. The investigations also included reviews of procedures, cultures, and microbiologic studies of infecting, contaminating, and colonizing strains. RESULTS: Sixty-two case patients were identified, 49 (79 percent) of whom underwent surgery during an epidemic period. Postoperative complications were more frequent during the epidemic period than before it. Only exposure to propofol, a lipid-based anesthetic agent, was significantly associated with the postoperative complications at all seven hospitals. In six of the outbreaks, an etiologic agent (Staphylococcus aureus, Candida albicans, Moraxella osloensis, Enterobacter agglomerans, or Serratia marcescens) was identified, and the same strains were isolated from the case patients. Although cultures of unopened containers of propofol were negative, at two hospitals cultures of propofol from syringes currently in use were positive. At one hospital, the recovered organism was identical to the organism isolated from the case patients. Interviews with and observation of anesthesiology personnel documented a wide variety of lapses in aseptic techniques. CONCLUSIONS: With the increasing use of lipid-based medications, which support rapid bacterial growth at room temperature, strict aseptic techniques are essential during the handling of these agents to prevent extrinsic contamination and dangerous infectious complications.

Immune response to intramuscular revaccination after primary intradermal vaccination against hepatitis B.

We studied the immune response to (re)vaccination with three 1-mL doses of recombinant hepatitis B vaccine administered intramuscularly on days 0, 30, and 180 to 75 public safety workers (PSWs) who had not developed antibody to hepatitis B surface antigen (anti-HBs) after three intradermal doses of hepatitis B vaccine; to 45 PSWs who had initially developed antibody but did not have detectable levels 11 months after intradermal vaccination; and to 16 hepatitis B-susceptible PSWs. Levels of anti-HBs were measured on days 14 and 210 after the first intramuscular dose. Overall, 46 (61%) of 75 PSWs in the initial-nonresponse group, 43 (96%) of 45 PSWs in the lost-response group, and 5 (31%) of 16 PSWs in the new-vaccinee group had anti-HBs titers of > or = 10 mIU/mL on day 14. On day 210 (after three doses), the figures were 62 (89%) of 70 PSWs in the initial-nonresponse group, 43 (98%) of 44 PSWs in the lost-response group, and 15 (94%) of 16 PSWs in the new-vaccine group. We conclude that persons who do not seroconvert after intradermal vaccination should receive three doses of hepatitis B vaccine by the intramuscular route.

Infectious complications associated with temporary epidural catheters.

To identify cases of deep-tissue or local infection associated with temporary epidural catheters, we reviewed medical records from 1980 through 1992 and prospectively followed up patients with temporary epidural catheters from January 1993 through June 1993 who were hospitalized at a large, tertiary referral hospital. We identified seven cases of temporary epidural catheter-associated infection, including three cases of deep-tissue infection (paraspinal muscle abscess, epidural abscess, and meningitis) and four cases of local infection. The infections were diagnosed 2 days to 22 days following insertion of the epidural catheter. Staphylococcus aureus was isolated in four (57%) of the cases. All seven patients were treated with removal of the catheter and antimicrobial therapy; three patients also required surgical drainage for management of deep-tissue infection. Our findings emphasize the importance of daily inspection and prompt removal of temporary epidural catheters if infection is suspected.

Possible nosocomial transmission of Pseudomonas cepacia in patients with cystic fibrosis.

OBJECTIVE: To determine whether nosocomial transmission of Pseudomonas cepacia occurred at a hospital with endemic P cepacia infection of patients with cystic fibrosis. DESIGN: Two retrospective case-control studies. SETTING: A large pediatric cystic fibrosis center. PARTICIPANTS: To assess risk factors for acquisition of P cepacia, 18 cases, defined as any patient with cystic fibrosis with first documented isolation of P cepacia in 1988 or 1989, were compared with 18 matched P cepacia-negative controls with cystic fibrosis. To assess potential modes of nosocomial P cepacia transmission, 14 cases with a hospitalization(s) between their last P cepacia-negative culture and first P cepacia-positive culture were compared with 14 hospitalized P cepacia-negative controls with cystic fibrosis. METHODS: Handwiping cultures (N = 68) and selective environmental cultures were performed. MAIN RESULTS: Cases tended to be more likely than controls to have been hospitalized at the cystic fibrosis center in the 3 months before their first P cepacia-positive culture (P = .08). In addition, cases tended to be more likely than hospitalized controls with cystic fibrosis to have had a P cepacia-positive roommate (P = .06) before becoming colonized with P cepacia organisms. Pseudomonas cepacia was cultured from the hands of two individuals: a P cepacia-colonized patient who had just undergone chest physiotherapy and consequent coughing and the investigator who shook the P cepacia-positive patient's hand after the patient's procedure. CONCLUSIONS: These results suggest that in this cystic fibrosis center, hospitalization is a risk factor for P cepacia acquisition and that person-to-person transmission of P cepacia may occur in the hospital via hand contact.

Prevalence of illicit drugs detected in the urine of women of childbearing age in Alabama public health clinics.

Each year, it is estimated that from 350,000 to 739,000 U.S. infants are exposed in utero to one or more illicit drugs. To estimate the prevalence of and risk factors for illicit drug use by women of childbearing age in Alabama, during 2 months in 1991 the authors collected patient-reported histories, clinical histories, and urine specimens from 6,195 women statewide attending public health maternity clinics, family planning clinics, and a high-risk referral obstetrical clinic. Blind drug screening of urine specimens for marijuana, cocaine, opiates, barbiturates, and amphetamines was performed with the use of a fluorescent polarization immunoassay. The overall prevalence of positive results for drugs tested was 10.1 percent, including 8.4 percent of the 3,554 pregnant and 12.3 percent of the 2,571 nonpregnant women screened. The drugs most frequently detected were marijuana and cocaine. Characteristics of the subjects associated with a higher prevalence of positive results for any drug tested or for marijuana included white race, older age, being divorced, non student occupation, having 12 or less years of education, attending a clinic located in a suburban county, self-reported substance use, increased risk for human immunodeficiency virus infection, and reproductive history. Characteristics of women with positive screening for cocaine results were similar to those who tested positive for any drug, except that the prevalence of cocaine was higher among black women and those attending urban county clinics and did not vary by years of education. Patient-reported histories of drug use were insensitive in identifying women who had positive drug screening results (sensitivity, 6.3 percent; specificity, 98.2 percent). Thus, in this study,the use of illicit drugs among women of childbearing age attending public clinics in Alabama was common and emphasizes the need for targeted drug education and interventions to reduce the impact of drug use on this high-risk population.

Epidemic gram-negative bacteremia in a neonatal intensive care unit in Guatemala.

BACKGROUND: Nosocomial bloodstream infection is an important cause of morbidity and mortality among neonates. From September 1 through December 5, 1990 (epidemic period), gram-negative bacteremia developed in 26 neonates after their admission to the neonatal intensive care unit (NICU) of Hospital General, a 1000-bed public teaching hospital in Guatemala with a 16-bed NICU. Twenty-three of the 26 patients (88%) died. METHODS: To determine risk factors for and modes of transmission of gram-negative bacteremia in the NICU, we conducted a cohort study of NICU patients who had at least one blood culture drawn at least 24 hours after admission to the NICU and performed a microbiologic investigation in the NICU. RESULTS: The rate of gram-negative bacteremia was significantly higher among patients born at Hospital General, delivered by cesarian section, and exposed to selected intravenous medications and invasive procedures in the NICU during the 3 days before the referent blood culture was obtained. During the epidemic period, the hospital's chlorinated well-water system malfunctioned; chlorine levels were undetectable and tap water samples contained elevated microbial levels, including total and fecal coliform bacteria. Serratia marcescens was identified in 81% of case-patient blood cultures (13/16) available for testing and from 57% of NICU personnel handwashings (4/7). Most S. marcescens blood isolates were serotype O3:H12 (46%) or O14:H12 (31%) and were resistant to ampicillin (100%) and gentamicin (77%), the antimicrobials used routinely in the NICU. CONCLUSIONS: We hypothesize that gram-negative bacteremia occurred after invasive procedures were performed on neonates whose skin became colonized through bathing or from hands of NICU personnel.