Effectiveness of a Multicomponent Intervention to Reduce Multidrug-Resistant Organisms in Nursing Homes

Study Population and Design

We conducted a cluster randomized clinical trial of recently admitted NH patients in Michigan from September 2016 to August 2018. Eligibility criteria included new admission to the NH and enrollment within 14 days and provision of written informed consent by the patient or their legally authorized representative. Exclusion criteria were non–English speaking and receipt of hospice care. Enrolled patients were followed-up for up to 6 months or discharge. The cluster randomized clinical trial design was chosen to avoid potential intervention contamination that occurs when randomizing patients^18,19 and to avoid the influence of secular trends.²⁰

Intervention

The multicomponent intervention included 6 factors (Table 1). The intervention was developed after a review of previously published studies in both hospital and nursing home settings. Briefly, the multicomponent intervention included enhanced barrier precautions for patients with high risk, chlorhexidine bathing, MDRO surveillance, environmental cleaning education and feedback, hand hygiene promotion, and health care worker (HCW) education and feedback.

Clinical Data Collection

Research personnel reviewed enrolled patient medical records to collect demographic and clinical data. Functional status was measured using the Physical Self-Maintenance Scale, which ranges from 6 (full independence) to 30 (full dependence). The 6 Physical Self-Maintenance Scale categories are bathing, dressing, feeding, ambulation, grooming, and toileting.²³ Open wounds were defined as having an open chronic wound that required regular dressing changes or a wound vacuum. This included purulent wounds, vascular ulcers, diabetic ulcers, open surgical wounds, or pressure ulcers greater than stage 1. We defined prolonged hospitalization before NH admission as longer than 14 days.⁶ The Charlson Comorbidity Index was used as a cumulative comorbidity measure.²⁴ Prior antibiotic exposure was defined as documented receipt of an antibiotic within 30 days before study enrollment.⁶ Clinical infections were objectively defined as the presence of both a clinical note in the patient’s medical record and the prescription of a systemic antibiotic.²² We reviewed the Minimum Data Set 3.0 to confirm the dates and destinations of patient discharges.²⁵

Microbiologic Methods

Specimens were collected using sterile swabs in transport media (BactiSwab; Remel) and assessed for MDROs including methicillin-resistant Staphylococcus aureus (MRSA), vancomycin-resistant enterococci (VRE), and resistant gram-negative bacilli (R-GNB) in our research laboratory using previously described, standard microbiologic methods and appropriate quality controls.⁶ We classified gram-negative bacilli as R-GNB if they were resistant to any of the following antibiotics: ceftazidime (30 μg), ceftazidime and clavulanic acid (30/10 μg), ciprofloxacin (5 μg), or imipenem (10 μg).⁶

Sample Size

Our sample included 6 privately-owned NHs with an mean (range) bed size of 107 (74-143) beds (eTable 2 in Supplement 2). Power calculations were based on the assumption of a 30% reduction in new MDRO acquisition as a result of our intervention. The desired statistical power was 0.80, and significance was set at 2-sided P = .05. We adjusted for intracluster correlation using an intraclass correlation coefficient of 0.07.²² We estimated that we would need to recruit approximately 239 patients over 2 years in both groups. Assuming a 50% consent rate, we anticipated screening approximately 478 patients to achieve our sample size. Thus, 3 NHs per study arm would suffice in reaching 80% power for our primary aim of reducing MDRO prevalence between the study arms.

Randomization and Masking

As randomizing participants or HCWs within NHs may lead to the transient crossover of intervention knowledge between subgroups and attenuation of the true effect of the intervention, we cluster randomized NHs. Facilities were randomized by their projected number of recruited participants using computer-generated randomization procedures by the study statisticians (M.K. and A.G.).²² While field staff (K.E.G. and B.J.L.) were aware of facility intervention status, the researcher processing microbiological specimens (M.C.) was masked.

Main Outcomes

The primary study outcome was the presence of patient MDRO colonization or contamination of the patient’s room environment over the duration of the study. The secondary outcome was the new acquisition of MDROs in patients. We also report an exploratory outcome, NH onset of clinically documented infections.

Statistical Analysis

To evaluate the impact of the intervention on the prevalence of any MDRO and each MDRO (MRSA, VRE, and R-GNB) for patient and environmental specimens separately, we used logistic regression with random effect, a type of generalized mixed-effect model. We performed unadjusted and adjusted analyses (adjusting for age, sex, race/ethnicity, functional status, and device use).

In addition to separate analyses, we performed a post hoc analysis of pooled patient and room environment data and analyzed them jointly to assess the effect of the intervention on MDRO presence in the patient-environment setting (eFigure 1 in Supplement 2). Our analyses included 245 patient-environment dyads, encompassing 1586 total visits. We adjusted for multilevel data by considering 2 random effects: dyads clustered within visits and patient vs environment within each dyad.

To estimate the impact of the intervention on MDRO acquisition, we analyzed data from patients with more than 1 visit who did not have any MDRO of interest at study enrollment. We used univariable and multivariable Cox regression modeling to estimate the time to new acquisition of each MDRO and a combined any new MDRO outcome. The multivariable Cox regression model adjusted for age, sex, race/ethnicity, functional status, and device use. Race/ethnicity was collected via health record review per National Institutes of Health requirements. In an exploratory analysis, we added patient room contamination as an additional risk factor for patient MDRO acquisition to the multivariable Cox regression models. The acquisition rate was defined as new acquisition events per 1000 patient-days. All regression analyses were clustered by NH. Although not powered to test for the intervention’s effect on clinically documented infections, we compared infection rates in each group using Fisher exact test.

Data were analyzed using Stata statistical software version 15.1 (StataCorp). Preliminary analysis was performed from November 2018 to April 2019. Analysis for first submission was performed from October 2019 to February 2020.

Study Group Characteristics

The mean (SD) age of all enrolled patients was 72.5 (13.6) years, 134 (54.7%) were women, and 132 patients (53.8%) were non-Hispanic White (Table 2). Nearly all patients (231 patients [94.3%]) were admitted for an anticipated short stay, and 235 patients (95.9%) were recorded as receiving postacute care. Patient clinical characteristics were generally balanced between the study arms. However, the intervention NHs had a lower proportion of non-Hispanic White patients than the control NHs (49 patients [43.4%] vs 83 patients [63.4%]). In the acquisition analysis sample, which was restricted to patients with at least 2 study visits, we also found that intervention patients had lower levels of functional disability than the control patients (eTable 3 in Supplement 2). We distributed 11 808 hand sanitizer bottles and 84 450 chlorhexidine wipes to the intervention facilities.

MDRO Prevalence

We collected 1556 patient specimens from intervention NH patients, of which 245 specimens (15.8%) had test results positive for 1 or more MDROs: 93 specimens (6.0%) with MRSA, 102 specimens (6.6%) with VRE, and 81 specimens (5.2%) with R-GNB (eTable 4 in Supplement 2). We collected 2098 control NH patient specimens, of which 434 specimens (20.7%) had test results positive for 1 or more MDROs: 169 specimens (8.1%) with MRSA, 208 specimens (9.9%) with VRE, and 151 specimens (7.2%) with R-GNB. By univariate analyses, the multicomponent intervention reduced the odds of the presence of any MDRO (odds ratio [OR], 0.72; 95% CI, 0.60-0.85), MRSA (OR, 0.73; 95% CI, 0.56-0.94), VRE (OR, 0.64; 95% CI, 0.50-0.82), and R-GNB (OR, 0.71; 95% CI, 0.54-0.94) in the patient.

We collected 2372 environmental specimens from rooms of patients in the intervention NHs. Of these, 309 specimens (13.0%) had test results positive for 1 or more MDRO: 122 specimens (5.2%) with MRSA, 152 specimens (6.4%) with VRE, and 66 specimens (2.8%) with R-GNB. We collected 3234 environmental specimens from rooms of control NH patients. Of these, 555 specimens (17.2%) had test results positive for 1 or more MDRO: 232 specimens (7.2%) with MRSA, 308 specimens (9.5%) with VRE, and 106 specimens (3.3%) with R-GNB. By univariate analyses, the multicomponent intervention reduced the odds of the presence of any MDRO in the environment (OR, 0.72; 95% CI, 0.62-0.84) and specifically of MRSA (OR, 0.70; 95% CI, 0.56-0.88) and VRE (OR, 0.65; 95% CI, 0.53-0.80) environmental contamination. There was no significant reduction in R-GNB environmental contamination (OR, 0.84; 95% CI, 0.62-1.15). MDRO surveillance data by study visit are presented in eFigure 2 in Supplement 2. In-depth, swab-level MDRO data are presented in eTable 5 and eTable 6 in Supplement 2.

For the primary outcome, after adjusting for patient-level covariates (ie, age, sex, race/ethnicity, functional status, and device use), there was no statistically significant difference between intervention and control patients in overall patient MDRO colonization (adjusted OR, 0.57; 95% CI, 0.29-1.14; P = .12). There was a statistically significant reduction in overall environmental MDRO contamination for intervention NHs (adjusted OR, 0.57; 95% CI, 0.35-0.94, P = .03) (Table 3).

When we used multilevel modeling to evaluate the intervention’s effect on the MDRO prevalence in the pooled samples from both the patient and their room environment, also controlling for age, sex, race/ethnicity, functional status, and device usage, we detected a statistically significant 42% reduction in overall MDRO prevalence in the intervention NH patients when compared with control NH patients and their environment (adjusted OR, 0.58; 95% CI, 0.35-0.98; P = .04). Among specific MDROs, we found no significant reductions in prevalence of MRSA (adjusted OR, 0.44; 95% CI, 0.16-1.20; P = .11), VRE (adjusted OR, 0.53; 95% CI, 0.27-1.02; P = .06), or R-GNB (adjusted OR, 0.67, 95% CI, 0.41-1.09; P = .11).

New Acquisitions of MDROs

The rates for a new MDRO acquisition were 36.3 (95% CI 27.2-48.3) infections per 1000 patient-days in the intervention NHs and 42.7 (95% CI 33.6-54.3) acquisitions per 1000 patient-days in the control NHs (Table 4). The rates of new MRSA acquisition were 2.4 (95% CI 0.9-6.4) acquisitions per 1000 patient-days in intervention NHs and 8.0 (95% CI 5.0-12.8) acquisitions per 1000 patient-days in control NHs. The rates of new VRE acquisition were 8.4 (95% CI 4.9-14.4) acquisitions per 1000 patient-days in intervention NHs and 9.0 (95% CI 5.6-14.4) acquisitions per 1000 patient-days in control NH patients. The rates of new R-GNB acquisition were 13.5 (95% CI 8.8-20.7) acquisitions per 1000 patient-days in intervention NHs and 13.9 (95% CI 9.5-20.2) acquisitions per 1000 patient-days in control NHs.

When adjusting for patient-level covariates, the intervention was not associated with significantly decreased time to new MRSA acquisition (hazard ratio [HR], 0.20; 95% CI, 0.04-1.09; P = .06), VRE acquisition (HR, 0.84; 95% CI, 0.46-1.53; P = .57), or R-GNB acquisition (HR, 1.14; 95% CI, 0.73-1.78; P = .57). These observations were conserved after the addition of patient room environment contamination to the multivariable Cox proportional hazard model.

Other Adverse Events

We identified 36 new clinically documented infections with onset after admission (eTable 7 in Supplement 2). The most common infections included urinary tract infections (12 infections [25.0%]), skin and soft tissue infections (7 infections [19.4%]), and lower respiratory infections (6 infections [16.7%]). We found no difference in clinically-documented infections by study group. There were 45 acute care rehospitalizations, including 19 (42.2%) at intervention NHs and 26 (57.8%) at control NHs, and there were 5 deaths, including 3 (60.0%) at intervention NHs and 2 (40.0%) at control NHs. These outcomes were considered unrelated to the study. There were no adverse events related to chlorhexidine bathing, including no reported adverse skin conditions or allergic events. Additionally, the intervention did not change the number of progress notes written in the patient’s medical records nor change the number of therapy sessions (eTable 8 in Supplement 2). Participants in both groups reported visiting the dining room, therapy, and hallway preceding their study visits, with no significant differences in activities identified across study groups (eTable 9 in Supplement 2) or hand contamination after select activities (eTable 10 in Supplement 2).

Strengths and Limitations

Our study has several strengths. First, this study is one of the more comprehensive multicomponent infection prevention interventions in the postacute care NH setting.^22,44,45 As such, this intervention contributes to the emerging literature on the potential of multicomponent interventions in this high-risk, MDRO-burdened health care environment. Second, the pragmatic study design and multicomponent intervention implementation accurately represent real-world infection prevention efforts, accounting for the persistent challenges inherent to these facilities. Additionally, our pragmatic cluster randomized clinical trial had a far stronger design than existing pre-post experimental studies of MDRO and infection prevention, most notably by addressing bias associated with secular time-trends, such as reductions in MDRO transmission due to ongoing, nonintervention infection prevention efforts.²⁰ Third, this study evaluated the intervention’s effect on patient room environment contamination.

Our study has several limitations. First, our results may not be generalizable to other regions and types of nursing facilities. It is possible that infection control and environmental cleaning practices varied between the sites; however, intervention sites received a standardized environmental cleaning education. Our intervention is resource-intensive and thus may require leadership and material support to implement. Second, future studies should extend our results to develop interventions that reduce new infections. Third, while analysis of individual organisms can assist in understanding intervention trends and mechanisms, the interpretation of the intervention’s effect on individual organisms was not the focus of the study and needs to be addressed in future work. Fourth, the enrollment rate in the intervention arm was slightly lower than the control arm, which could reflect reduced patient desire to participate in intervention procedures immediately after an acute care hospitalization. Fifth, while studies have shown the efficacy of infection prevention interventions in NHs, it is difficult to assess the effect of individual intervention components.⁴⁵