The coronavirus disease 2019 (COVID-19) pandemic has strained healthcare organizations and disrupted delivery of care worldwide. To accommodate the surge of COVID-19 patients, healthcare facilities increased critical care capacity, reduced elective and nonemergency procedures, and modified staffing practices, and modified the use of personal protective equipment, among other strategies. Reference Baker, Sands and Huang1–Reference Weiner-Lastinger, Pattabiraman and Konnor8 Shortages of staff, personal protective equipment, and supplies (eg, ventilators and disinfectants) have been reported during the pandemic, along with overburdening of healthcare workers and changes in infection control guidance. Reference Cheeyandira2,Reference Rebmann, Alvino and Holdsworth7,Reference Gohar, Larivière and Nowrouzi-Kia9,Reference Rebmann, Alvino, Mazzara and Sandcork10 Additional infection prevention measures, such as universal masking of patients, were also introduced to minimize severe acute respiratory coronavirus virus 2 (SARS-CoV-2) transmission within facilities. Reference Wang, Ferro, Zhou, Hashimoto and Bhatt11 All of these changes may have influenced the effectiveness of infection prevention and control of healthcare-associated infections (HAIs).
Several studies have reported a rising incidence of various HAIs since the onset of the COVID-19 pandemic. Baker et al Reference Baker, Sands and Huang1 described the following increases, all of which correlated to surges of SARS-CoV-2 infections: 43% in catheter-associated urinary tract infections, 60% in central-line–associated bloodstream infections, 44% in methicillin-resistant Staphylococcus aureus (MRSA) bacteremia, and 24% in multidrug-resistant organism (MDRO) bacteremia overall. Reference Baker, Sands and Huang1 Similarly, in 2021, Weiner-Lastinger et al Reference Weiner-Lastinger, Pattabiraman and Konnor8 reported increases in the national-level standardized infection ratios for central-line–associated bloodstream infections, catheter-associated urinary tract infections, ventilator-associated events, and MRSA bacteremia in the 2020 National Healthcare Safety Network. Reference Weiner-Lastinger, Pattabiraman and Konnor8 Alternatively, some hospitals have reported reductions in HAIs, possibly due to enhanced infection control practices and heightened awareness of proper personal protective equipment use. Reference Bentivegna, Luciani, Arcari, Santino, Simmaco and Martelletti12,Reference Cole and Barnard13 Overall, previous research has focused primarily on reporting coinfections and superinfections in COVID-19 hospitalized patients. Reference Kubin, McConville and Dietz14 However, whether non–COVID-19 patients were also at a greater risk of HAIs during the pandemic remains unclear.
We sought to quantify whether COVID-19 pandemic changed the risk for acquiring a multidrug-resistant HAI for both COVID-19 and non–COVID-19 hospitalized patients in a multicenter retrospective cohort. We used admission and clinical data from 4 hospitals to construct mixed-effect logistic regression models and to quantify patient-level risk of acquiring an HAI associated with 1 of 8 multidrug-resistant pathogens: Acinetobacter spp, Clostridioides difficile, Enterococcus faecium, Escherichia coli, Klebsiella pneumoniae, Pseudomonas aeruginosa, Staphylococcus aureus, and general MDRO Enterobacteriaceae spp. We estimated the effect of the COVID-19 pandemic period on pathogen-specific HAI acquisition risks for hospitalized patients relative to a prepandemic period. We also quantified the effects of COVID-19 diagnoses preceding HAI onset and intrahospital COVID-19 burden during patients’ hospital stays on individual HAI risk during the pandemic period.
Methods
Study population
A multicenter retrospective cohort study was conducted in 4 hospitals in the St. Louis area (see Appendix 1 for detailed hospital characteristics). Admission and clinical data were collected for patients admitted between January 2017 and August 2020, discharged no later than September 2020, and hospitalized ≥48 hours. When discharged patients were readmitted to the same hospital <24 hours later, hospitalization data were concatenated into a single record. The data set consists of 260,055 unique admissions associated with 157,120 different patients. We extracted clinical and demographic data for this cohort from the BJC Medical Informatics database, including laboratory and microbiological test results, antibiotic use, and days spent in intensive care units (ICUs). Comorbidity information for this cohort was obtained using International Classification of Disease, Tenth Revision (ICD-10) codes with Elixhauser groupings. 15,Reference Elixhauser, Steiner, Harris and Coffey16 Cohort members were hospitalized for 6.83 days on average (SD, 8.30 days).
Patients with a positive culture from blood, urine, or respiratory tract specimen for any of 5 MDRO groups previously defined by the Centers for Disease Control and Prevention and described by Burnham et al Reference Burnham, Olsen, Stwalley, Kwon, Babcock and Kollef17 were identified: MDRO Enterobacteriaceae spp, S. aureus, P. aeruginosa, Acinetobacter spp, and E. faecium. See Appendix 2 for MDRO group definitions. C. difficile infection was identified by positive toxin enzyme immunoassay from stool. Within the MDRO Enterobacteriaceae spp, we also separately analyzed multidrug-resistant E. coli and K. pneumoniae. We categorized infections as hospital-onset HAIs if they were first identified ≥48 hours following patient admission. 18 COVID-19 status during admissions were based on results of SARS-CoV-2 PCR testing and/or a diagnosis code for COVID-19 (ICD-10-CM code U07.1). Medical record review was performed for admissions with a COVID-19 ICD-10-CM diagnosis code without a positive laboratory result to confirm a clinical diagnosis. To calculate the infection pressures associated with COVID-19 and HAIs, admissions for the first and last 24 days were removed. The final number of admissions used in the statistical analysis were 250,539.
Statistical analysis
We constructed independent mixed-effects logistic regression models to estimate patients’ individual-level risk of infection with the pathogens of interest during hospitalization. Mixed-effects logistic regression models are among methods often used to characterize individual-level infection risk while considering that patients are nested within hospital.
Reference Gomila, Carratalà and Eliakim-Raz19,Reference Clements, Tong, Morton and Whitby20
All analyses were done using the lme4 R package version 1.1-27.1
Reference Bates, Mächler, Bolker and Walker21
in R Studio version 1.4.1717 running R version 4.0.3 software. We carried out 2 sets of regression analyses, which we refer to as the “full period” and “COVID-19 period” model sets, and we describe predictor variables included in each model set in Table 1. These model sets were intended to answer 2 distinct questions. The former set identified significant (P ≤ .05) differences in patient-level HAI risk associated with the initial 6 months of the COVID-19 pandemic. We used “COVID-19 period” models to determine whether patients with a COVID-19 diagnosis were more or less likely to acquire these HAIs than those not diagnosed with COVID-19, and whether hospital-level COVID-19 burden modulated HAI risk for both COVID-19 and non–COVID-19 admissions. The fixed effects of interest in our models were COVID-19_period, COVID-19_diagnosis
ix
, and the hospital-specific COVID-19 burden during a patient stay (
${\Delta _{h{\tau _{ix}}}}$
). The following variables were also included as confounders: time spent in the hospital prior to observed infections, the relative amount of time spent on antibiotics and in intensive care units, and pathogen-specific infection pressures acting on admitted patients. See Appendix 3 (online) for additional information on our model-fitting and model selection procedures.
Table 1. Description of Predictor Variables Included as Fixed Effects in Mixed-Effect Logistic Regression Models
Note HAI, healthcare-associated infection; ICU, intensive care unit; LOS, length of stay.
This study was approved by the Washington University School of Medicine Institutional Review Board (IRB ID no. 202010057) with a waiver of informed consent.
Results
The numbers of admissions readily decreased during the early phase of the pandemic, followed by an increase in admissions, to numbers more similar to ranges observed during preceding years (Fig. 1). Figure 2 shows the temporal trends for the estimated infection pressures, calculated as the proportion of patient days contributed by patients with a confirm diagnosis for a given MDRO or COVID-19. There is no clear change in the trends of the MDRO infection pressures in the COVID-19 pandemic period (Fig. 2a). The pandemic period was, however, associated with a notable increase in the proportion of admitted patients presenting with SARS-CoV-2 infection during the hospital stay (Fig. 2b).
Figure 1. Sum admissions with lengths of stay ≥2 days admitted to study hospitals each week. Weeks containing or following the first observed COVID-19 patient are highlighted in blue and begin at week 0. Weeks preceding COVID-19 patient admissions are shown were assigned negative IDs.
Figure 2. Estimated infection pressures for healthcare-associated infection (HAI) pathogens of interest (a) and COVID-19 burden (b) calculated for each individual admission record. Median estimates are shown by the black line, and weeks containing or following the first observed COVID-19 patient are highlighted in blue. All COVID-19 burden estimates for weeks preceding “Week −49” are equal to 0 and are not shown.
Adjusted odds ratios associated with COVID-19–period effects on HAI occurrence are given in Table 2. We observed significant COVID-19–period effects for Acinetobacter spp, P. aeruginosa, and overall Enterobacteriaceae spp HAIs, for which the odds ratios suggest hospital-onset infection probabilities increased by 2.64 (95% confidence interval [CI], 1.22–5.73) times, 1.44 (95% CI, 1.03–2.02) times, and 1.25 (95% CI, 1.00–1.58) times relative to prepandemic time points, respectively. The frequency of respiratory and urine cultures for patients remained unchanged between the prepandemic period and the pandemic period. Also, 5% of patients had a respiratory culture taken regardless of the period. Rates of urine culture were also very similar between the periods: 16% during the prepandemic period versus 17% during the pandemic period.
Table 2. Estimated Effects of the “COVID-19_Period” Variable on Individual-Level MDRO HAI Risk
Note. MDRO, multidrug-resistant organism; HAI, healthcare-associated infection; OR, odds ratio; CI, confidence interval.
a Adjusted ORs denoting the relative risk of pathogen-specific hospital-onset infections during COVID-19 period admissions are given. ORs for other model variables are not shown. Wald 95% CIs are listed in parentheses.
b A reduced or alternate version of the “full period” model was fit to the data.
Within the pandemic period, COVID-19 patients had both urine and respiratory cultures taken more frequently compared to non–COVID-19 patients. Also, 16% of COVID-19 patients had respiratory cultures versus 4% of non–COVID-19 patients during the pandemic period. Furthermore, 27% of COVID-19 patients had urine cultures versus 16% of non–COVID-19 patients during the pandemic period. Table 3 lists odds ratios associated with effects of COVID-19 diagnoses immediately preceding HAIs and hospital COVID-19 burden (
${\Delta _{h{\tau _{ix}}}}$
) during patient stays. Overall, a prior COVID-19 diagnosis and the hospital COVID-19 burden were not associated with an increase in patient-level risk for MDRO infection. The only exception was that COVID-19 patients were 4.18 (95% CI, 1.98–8.81) times more likely to acquire a hospital-onset MDRO S. aureus infection. Interestingly, half of our models estimated that COVID-19 burden had protective effects on infection probabilities (ie, odds ratio < 1); however, these effects were not significant at an α level of 0.05. Estimates for all predictors included in all the models are reported in Appendix 3 (online). Overall, number of prior days in the ICU was a significant variable for all HAI risk models, and prior days of antibiotics was also significant for most models.
Table 3. Estimated Effects of COVID-19 Diagnoses and COVID-19 Burden (Δ
${_{ih\tau }}$
) on Individual-Level MDRO HAI Risk
Note. MDRO, multidrug-resistant organism; HAI, healthcare-associated infection; OR, odds ratio; CI, confidence interval.
a Adjusted ORs denoting the relative risk attributable to 1-unit increases of these variables on pathogen-specific hospital-onset infection probability are given in the Table. ORs for other model variables are not shown. Wald 95% CIs are given in parentheses.
b A reduced or alternate version of the “COVID-19 period” model was fit to the data.
Discussion
We constructed mixed-effect logistic regression models to estimate COVID-19 effects on the patient-level risk for acquiring 8 MDRO HAIs. Patients admitted to hospitals between March and August 2020 (ie, during the COVID-19 pandemic period) were more likely to become infected with MDRO P. aeruginosa and Enterobacteriaceae spp and Acinetobacter spp than their prepandemic counterparts, but we did not observe significant period effects for other MDRO HAIs. The increase on infections with these pathogens was not linked to an increase in the frequency of culture during the pandemic period or COVID-19 diagnosis. However, MDRO P. aeruginosa and Enterobacteriaceae spp and Acinetobacter spp infections were associated with the preceding duration of stay in the ICU, suggesting that increased risk to infection may reflect a sicker population during the COVID-19 pandemic period. Prior studies have associated the increased HAI risks with a sicker patient population requiring ICU care, more frequent use of invasive devices, longer courses of antibiotics, and longer stays. 22 In our study, patients during the prepandemic period and non–COVID-19 patients in the pandemic period were similar, with comparable lengths of stays, age, comorbidities, and hospital mortality rates. However, there was a slightly increase in the percentage of ICU admissions for non–COVID-19 patients in some hospitals during the pandemic period (Table A1-2). Overall, COVID-19 patients were more severely ill, with longer lengths of stay, a higher ICU admission rate, and a higher hospital mortality rate (Table A1-2). In our COVID-19–period models, we accounted for the effects of patient susceptibility by including COVID-19 diagnosis and preceding days on antibiotics and in the ICU (Table 1). Additionally, we included intrahospital COVID-19 burden during the patient’s hospital stay as a variable to account for the potential indirect effects related to the burden of COVID-19 patients in the facilities, such as staff shortage and burnout or infection control changes (Table 1). Proportion of prior days in the ICU was a significant variable for all the HAI risk models, and proportion of prior days on antibiotics was also significant for most models (Appendices 2 and 3). However, COVID-19 burden was not significant in any of the HAI models (Table 3). Statistical models by Baker et al Reference Baker, Sands and Huang1 and standardized infection ratios by Weiner-Lastinger et al Reference Weiner-Lastinger, Pattabiraman and Konnor8 suggest that local COVID-19 burden was positively correlated with hospital-onset infections at the hospital level. These studies did not quantify the effects of COVID-19 burden at the individual level of risk, however. Our findings suggest that the signal observed in previous studies may capture effects of COVID-19–mediated changes in admitted patients’ hospital stay characteristics (eg, length of stay in ICUs) rather than facility-mediated COVID-19 infection effects, such as staff or infection control changes.
Our findings also indicated that having a positive COVID-19 diagnosis increased the probability of subsequently developing a hospital-onset MDRO S. aureus infection, but we could not conclude the same for other HAIs. In general, previous studies have indicated that COVID-19 patients are at increased risk, relative to non–COVID-19 patients, for hospital-onset MDRO HAIs. 22–Reference Martinez-Guerra, Gonzalez-Lara and de-Leon-Cividanes28 Superinfection with MDRO S. aureus has been described in previous reports describing HAI incidence in COVID-19 patients. Reference Weiner-Lastinger, Pattabiraman and Konnor8,Reference Kubin, McConville and Dietz14 Superinfection with MDRO S. aureus have been directly attributed to the management of COVID-19 patients, which often included used of catheters, intubation, and corticosteroid use.
Although we did not report an increased risk on other HAIs with COVID-19 diagnosis, others have reported increases in other HAIs. Garcia-Vidal et al Reference Garcia-Vidal, Sanjuan and Moreno-García26 noted that E. coli and P. aeruginosa infections were common in COVID-19 patients. Baskaran et al Reference Baskaran, Lawrence and Lansbury24 identified E. coli and K. pneumoniae as representing the majority of COVID-19–related HAIs. Study populations in these previous studies were comprised solely of critically ill COVID-19 patients or COVID-19 patients admitted to ICUs. Reference Baskaran, Lawrence and Lansbury24,Reference Blonz, Kouatchet and Chudeau25,Reference Grasselli, Scaravilli and Mangioni27 Thus, the distribution of HAIs affecting these populations may differ from ours. Our data set contained all ≥48-hour hospital admissions, and we did not limit our analyses to ICU admissions.
The main strengths of our study are the multisite nature of the study and that our cohort represented all hospitalized patients, including non–COVID-19 and non-ICU patients. The primary limitation of this study is that the scope of our data were restricted to patient-level post–hospital-admission attributes. Because we were unable to characterize hospital staffing and infection control policy changes, we ultimately lacked the ability to conclusively identify specific causes of COVID-19–related HAI risk modifications. Additionally, we defined the beginning of the COVID-19 period as the time point when the first recorded COVID-19 patient was admitted to one of the hospitals in this study. It is possible that patients infected with SARS-CoV-2 were admitted prior to this date, but their condition went undiscovered due to lack of testing and relative obscurity of the disease.
Our findings suggest that the COVID-19 pandemic has indirectly affected MDRO pathogen transmission in hospitals and that COVID-19 diagnosis may be related to MDRO S. aureus infection risk. These findings can be used to calibrate mechanistic transmission models for simulating HAI incidence, like that presented by Stephenson et al, Reference Stephenson, Lanzas and Lenhart29 and can help to illuminate causative HAI transmission mechanisms in this system. Our work will contribute to the development of effective interventions for preventing HAIs and MDRO transmission in hospitals.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/ice.2023.83
Acknowledgments
Financial support
This work was supported by the Centers for Disease Control and Prevention (grant nos. U01CK000587-01M001 and 75D30120C09598). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests
All authors report no conflicts of interest relevant to this article.
Open access
