Author + information
- Received April 23, 2014
- Revision received September 18, 2014
- Accepted September 19, 2014
- Published online January 6, 2015.
- Giampaolo Greco, PhD, MPH∗∗ (, )
- Wei Shi, MS∗,
- Robert E. Michler, MD†,
- David O. Meltzer, MD, PhD‡,
- Gorav Ailawadi, MD§,
- Samuel F. Hohmann, PhD‖,
- Vinod H. Thourani, MD¶,
- Michael Argenziano, MD#,
- John H. Alexander, MD∗∗,
- Kathy Sankovic, RN††,
- Lopa Gupta, MPH∗,
- Eugene H. Blackstone, MD††,
- Michael A. Acker, MD‡‡,
- Mark J. Russo, MD§§,
- Albert Lee, PhD‖‖,
- Sandra G. Burks, RN§,
- Annetine C. Gelijns, PhD∗,
- Emilia Bagiella, PhD∗,
- Alan J. Moskowitz, MD∗ and
- Timothy J. Gardner, MD¶¶
- ∗International Center for Health Outcomes and Innovation Research (InCHOIR), Department of Population Health Science and Policy, Icahn School of Medicine, Mount Sinai Medical Center, New York, New York
- †Department of Cardiothoracic Surgery, Montefiore Medical Center/Albert Einstein College of Medicine, New York, New York
- ‡Department of Medicine, University of Chicago, Chicago, Illinois
- §Division of Thoracic and Cardiovascular Surgery, University of Virginia School of Medicine, Charlottesville, Virginia
- ‖University HealthSystem Consortium, Chicago, Illinois
- ¶Clinical Research Unit, Division of Cardiothoracic Surgery, Emory University School of Medicine, Atlanta, Georgia
- #Division of Cardiothoracic Surgery, Department of Surgery, College of Physicians and Surgeons, Columbia University, New York, New York
- ∗∗Division of Cardiovascular and Thoracic Surgery, Department of Surgery, Duke University Medical Center, Durham, North Carolina
- ††Department of Thoracic and Cardiovascular Surgery, Cleveland Clinic Foundation, Cleveland, Ohio
- ‡‡Department of Surgery, Division of Cardiovascular Surgery, University of Pennsylvania School of Medicine, Philadelphia, Pennsylvania
- §§Barnabas Heart Hospital, Newark, New Jersey
- ‖‖Division of Cardiovascular Sciences, National Heart, Lung and Blood Institute, Bethesda, Maryland
- ¶¶Center for Heart and Vascular Health, Christiana Care Health System, Newark, Delaware
- ↵∗Reprint requests and correspondence:
Dr. Giampaolo Greco, Icahn School of Medicine at Mount Sinai, One Gustave L. Levy Place, Box 1077, New York, New York 10029.
Background Health care–associated infections (HAIs) are the most common noncardiac complications after cardiac surgery and are associated with increased morbidity and mortality. Current information about their economic burden is limited.
Objectives This research was designed to determine the cost associated with major types of HAIs during the first 2 months after cardiac surgery.
Methods Prospectively collected data from a multicenter, observational study of the Cardiothoracic Surgery Clinical Trials Network, in which patients were monitored for infections for 65 days after surgery, were merged with related financial data routinely collected by the University HealthSystem Consortium. Incremental length of stay (LOS) and cost associated with HAIs were estimated using generalized linear models, with adjustments for patient demographics, clinical history, baseline laboratory values, and surgery type.
Results Among 4,320 cardiac surgery patients (mean age: 64 ± 13 years), 119 (2.8%) experienced a major HAI during the index hospitalization. The most common HAIs were pneumonia (48%), sepsis (20%), and Clostridium difficile colitis (18%). On average, the estimated incremental cost associated with a major HAI was nearly $38,000, of which 47% was related to intensive care unit services. The incremental LOS was 14 days. Overall, there were 849 readmissions; among these, 8.7% were attributed to major HAIs. The cost of readmissions due to major HAIs was, on average, nearly threefold that of readmissions not related to HAIs.
Conclusions Hospital cost, LOS, and readmissions are strongly associated with HAIs. These associations suggest the potential for large reductions in costs if HAIs following cardiac surgery can be reduced. (Management Practices and the Risk of Infections Following Cardiac Surgery; NCT01089712).
Patients undergoing cardiac surgery are at risk for major postoperative infections (1–3), which carry devastating, if not fatal, clinical consequences and substantial costs, as reflected by prolonged hospitalizations and more frequent readmissions (4,5). Reducing the risk for health care–associated infections (HAIs) is a key priority for improving surgical care. HAIs are of increasing concern given that patients undergoing cardiac surgery are older and have multiple comorbidities, which further increase their infection risk. Some HAIs have recently decreased due to the identification and successful implementation of best practices. For example, in the United States, the rate of catheter-related bloodstream infections in the intensive care unit (ICU), decreased by 46% between 2008 and 2012, and the rate of surgical-site infections after cardiac procedures decreased by 30% over the same period (6–8). On the other hand, other important infections, such as pneumonia and sepsis, remain common, and other, previously uncommon infections, such as Clostridium difficile colitis, have been increasing in recent years (9,10). Although significant achievements have been made, there remains a nearly 5% risk for major post-operative infections in patients in the first 2 months following cardiac surgery, which is associated with a 10-fold higher risk for mortality (10).
Health care payers and policymakers have developed mechanisms aimed at preventing HAIs, including denial of reimbursement for the extra cost of treating HAIs considered preventable, mandatory public reporting of institutions’ HAI rates, and transparency regarding the level of adherence to national quality measures (11,12). However, detailed data regarding the economic impact of postoperative complications, particularly complications related to infectious processes, remain scarce. Measuring the economic impact of HAIs—especially in the context of invasive cardiac procedures, during which a patient is particularly vulnerable to infection—is essential for understanding the contribution of HAIs to rising health care costs and for developing sustainable approaches to preventing infection.
Research on the cost of HAIs has primarily focused on specific infection sites or pathogens, or has been limited to the index hospitalization. Few studies capture the full economic impact of a broad range of possible HAIs or the impact of these infections on hospital readmissions (13,14). Those studies that have addressed the economic impact of a broader spectrum of HAIs in cardiac surgery patients have typically relied on billing datasets (5,15), which have important limitations in identifying HAIs due to their reliance on International Classification of Diseases-Ninth Revision (ICD-9) infection codes (16,17). Only one in four HAIs, as detected by ICD-9 codes, meets standard Centers for Disease Control and Prevention (CDC)/National Healthcare Safety Network definitions and criteria (17). Billing data alone may lack the degree of clinical detail needed for adequately adjusting for patients' baseline clinical status. This study, however, combines two data sources: 1) clinical data from a prospective, multicenter, observational cohort study that evaluated the occurrence of major HAIs; and 2) economic data obtained from the University HealthSystem Consortium (UHC), and it addresses the economic burden of the major types of HAIs acquired within 65 days after cardiac surgery.
Between February 2010 and October 2010, a prospective, multicenter, observational study (n = 5,158) was conducted by the National Heart, Lung, and Blood Institute–sponsored Cardiothoracic Surgical Trials Network (CTSN) (trial investigators are listed in the Online Appendix) to assess the occurrence of HAIs after cardiac surgery. The study protocol was approved by the institutional review board at each participating clinical site and was compliant with the Health Insurance Portability and Accountability Act. All adult (≥18 years of age) cardiac surgery patients without active infections on admission were eligible to participate. Recruitment of all eligible patients continued until a pre-specified number of infectious events was reached. Data collected from this cohort were used in the present study and included demographics, body habitus, baseline laboratory results, comorbid conditions, type of surgery performed, infections, and readmissions within 60 ± 5 days of surgery. Patients had a follow-up visit or phone call at 30 and 60 ± 5 days after surgery for the collection of information about their health status and hospital readmission history. Discharge summaries and medical records were obtained and reviewed to determine the reason for the readmission in patients hospitalized at out-of-network institutions.
An independent committee of infectious-diseases experts, using definitions adapted from the CDC/National Healthcare Safety Network, reviewed and adjudicated all major postoperative infections. The last infectious event was reviewed in June 2012. The quality and completeness of the data were monitored as described elsewhere (10). Clinical data collected in the study were linked to patient-level economic data obtained directly from the sites or from the UHC, which is an alliance of U.S. academic medical centers. Data linkage was done using a combination of variables, including sex, date of birth, procedure date and type, admission and discharge dates, and hospital identification number. In addition to those from the index hospitalization, data up to 65 days after surgery were extracted from the UHC database, including cost data, revenue codes, and ICD-9 codes. Costs were obtained by multiplying charges by the cost center–specific cost-to-charge ratios for each institution. Such ratios are based on the annual Medicare cost reports submitted by individual hospitals. This method of approximating cost is widely used and provides reasonably accurate estimates of actual costs (18). From out-of-network readmissions (readmissions in hospitals not participating in the study), only length of stay (LOS) was available. Because the CTSN study protocol did not include readmissions for rehabilitation and emergency department visits, these types of events were filtered from UHC using the ICD-9 code for rehabilitation, the revenue code for emergency department visits, and direct verification with the clinical site.
Of the 10 centers participating in the CTSN study (9 in the United States and 1 in Canada), only the U.S. centers were included in this study, to use more homogeneous cost data. Of 4,614 patients undergoing cardiac surgery in the United States, data from 4,320 (93.6%) were matched to their corresponding financial records. Data from patients who could not be matched (6.4%) were distributed across all hospitals; these patients did not differ in baseline characteristics, including demographics, laboratory values, and comorbidities, from patients who could be matched. Data from unmatched patients were excluded from the analysis.
Endpoints and analysis
The endpoints for this study were incremental hospital LOS and infection cost.
Cost during the index hospitalization
Direct hospital costs associated with major HAIs were calculated separately for the index hospitalization and for subsequent readmissions. For the index hospitalization, extra costs associated with major HAIs were estimated using a generalized linear model (GLM), with a log-link function and a gamma distribution. This method adjusts for patient-related confounders while accounting for the nature of cost data, which are often skewed to the right with heteroskedasticity. Factors found to be associated with cost (at p < 0.15) were assessed in the multivariate analysis. To generate the final multivariate model, we removed statistically nonsignificant variables and refitted the model until all variables in the model had a p value of 0.05 or less. The final model included the baseline factors age; sex; race/ethnicity; body mass index; white blood cell count; hemoglobin and creatinine concentrations; ejection fraction; statuses of pulmonary disease, diabetes, hypertension, hypercholesterolemia, congestive heart failure, and peripheral vascular disease; history of cerebrovascular accident, cardiac surgery, and use of corticosteroids; and type of procedure and its duration. Covariates and parameter estimates are shown in Online Table 1.
The incremental cost—the additional cost associated with major HAIs—was then calculated using the recycled prediction method (19). Put simply, via the parameters estimated through the GLM approach, we estimated the incremental cost due to major HAIs by varying the infection status while the other parameters were held constant. The mean difference of the two predictions—with and without major HAI—provided the incremental cost attributable to major HAI. Standard errors and confidence intervals (CIs) were derived by 1,000 bootstrap resampling runs. In each run, we randomly drew patients with replacement from each group separately (infection and noninfection), thereby creating 1,000 pairs of bootstrap samples. GLM modeling was repeated for each pair to get the estimated predicted mean differences. By ordering all of the 1,000 random estimates, the 2.5 and 97.5 percentiles were used as the confidence limits.
Cost of rehospitalizations
Hospital readmissions were stratified into two categories, depending on whether their occurrence was attributable to HAI or not. Descriptive statistics were then calculated. Cost data were available only for readmissions to hospitals participating in the CTSN study.
Length of stay
Duration of hospital stay was obtained from all index hospitalizations and all readmissions. For the index hospitalizations, incremental LOS was calculated using GLM with a log-link and a gamma distribution, and adjustment for the same variables used in the cost model. For rehospitalizations, the incremental LOS associated with infections was considered equivalent to the mean LOS of the rehospitalizations attributed to infections.
All analyses were conducted utilizing SAS version 9.2 (SAS Institute Inc., Cary, North Carolina). Descriptive analyses were performed using the Wilcoxon-Mann-Whitney test for all continuous variables, and using the chi-square or Fisher exact test for all categorical variables.
Characteristics of the study population
Among the 4,320 patients in this cohort, the mean age was 64 ± 13 years; 66% were male; and the mean body mass index was 28 ± 6 kg/m2 (Table 1). Among patients with major HAIs, there were higher prevalences of prior cardiac surgery, congestive heart failure, hypertension, and history of stroke. Patients with major HAIs had lower ejection fractions and lower levels of hemoglobin, and marginally higher levels of creatinine. Major infections were more common in patients who had longer, urgent, or emergent surgeries. Transplant recipients and patients who received a ventricular assist device (VAD) were more likely to have developed major HAIs during the index hospitalization compared with patients who underwent other types of cardiac procedures (15.4% vs. 2.4%; p <0.0001).
During the course of the index hospitalization and of ensuing readmissions, 250 major HAIs developed in 194 patients (4.5% of the cohort). Of these, 119 patients (2.7%) acquired one or more major HAIs during their index hospitalization, and 88 patients (2%) had a major HAI associated with their readmission (Table 2). Among patients with HAI-related readmissions, 9 (10.2%) had received a transplant or VAD during the index hospitalization. The most common type of major HAI was pneumonia, followed by bloodstream infection, C. difficile colitis, and surgical-site infections.
Cost of the index hospitalization
On average, patients who developed a major HAI during the index hospitalization had a longer LOS than did patients who did not (mean: 33 vs. 9 days). Patients with infections also had a more costly hospital stay than did noninfected patients (mean: [2010 U.S. dollars] $110,155 vs. $31,530). During the course of a hospitalization, the level of resource utilization changed daily. For patients without major HAIs, the mean cost of care per day peaked at the first 2 days around $5,000, and then declined sharply, leveling off at about $2,000 per day within a week (Central Illustration). In contrast, in patients who acquired a major HAI, the mean cost per day remained sustained for a longer period of time, gradually declining in roughly 3 weeks to a level comparable to that of noninfected patients. After adjustment for patient-related confounders, the additional cost per major HAI amounted to almost $38,000 (Table 3). On average, the LOS of patients who had a major HAI was 14 days longer than that of patients who did not acquire a major HAI. The adjusted incremental costs of major infections were $37,513 for the entire cohort and $39,264 when VAD and transplant recipients were excluded (Table 3). ICU expenditures constituted almost one-half of the total incremental cost, whereas hospital supplies, laboratory, and pharmacy costs together contributed about one-third of the extra cost (Figure 1).
Cost of rehospitalizations
In this cohort, 657 readmissions (15% of the index hospitalizations) occurred within 30 days. Sixty of these readmissions (9.1%) were attributed to major infections. Because patients were monitored for a period of 60 ± 5 days after surgery, we were able to observe that the high incidence of readmissions persisted beyond the conventional 30-day cutoff. During the entire follow-up period, after excluding rehabilitations and visits to the emergency department, there were 849 readmissions (19.7% of the index hospitalizations), of which 545 (64%) were to the same hospital where the initial surgery was done. Among these readmissions, 137 (16.1%) were infection related (including major and minor infections, e.g., urinary tract infections and superficial wound infections), and 74 (8.7%) were attributed to major HAIs alone. Readmissions due to a major HAI had 2.6-fold higher costs than did readmissions due to other causes, and their LOS was twice the LOS of readmissions unrelated to a major HAI (Table 4). Moreover, patients who had a major HAI during the index hospitalization had a higher rate of all-cause readmissions compared with that of patients who did not experience any major HAI at index (33% vs. 20%). On the basis of the observed difference in the rate of readmissions between these groups, we estimated one extra readmission for every 10 major HAIs developed during the index hospitalization. The corresponding extra cost per major HAI at index is reported in Table 4.
HAIs are a large impediment to achieving the full benefits of modern medicine in that they affect 1.7 million patients each year and are associated with nearly 100,000 deaths in the U.S. (20). In cardiac surgery patients, HAIs are the most common noncardiac complication and have been associated with increased morbidity and mortality, prolonged hospitalizations, and higher costs (4,5). Over the past decade, the field of cardiac surgery has undergone profound changes that have likely affected the burdens, clinical and financial, brought about by HAIs. These changes have stimulated the need for accurate, up-to-date evidence of such burdens. For example, a general trend toward shorter LOS, and a corresponding shift of care toward nonacute care facilities and patients’ homes (21), have partially shifted resource use to outpatient settings and readmissions. Studies on the effects of HAI, however, have almost exclusively focused on the index hospitalization, analyzing, with only a few exceptions, the occurrence of surgical-site infections after discharge (22,23). Additionally, patients undergoing cardiac surgery are increasingly older and affected by comorbidities such as diabetes and obesity (24). These demographic and epidemiological changes, along with modifications of appropriateness criteria (25,26), have reshaped the characteristics of the patient population undergoing cardiac surgery, and, therefore, the potential consequences of HAIs on both clinical and economic outcomes.
Among the type of major HAIs encountered in this study, during both index hospitalizations and readmissions, pneumonia was by far the most frequent (48% of major HAIs), followed by bloodstream infections (21%) (Table 2). These data are consistent with other reports from cardiac surgery and from ICU patients (4,27). Over the past decade, quality-improvement efforts have gained ground against HAIs such as catheter-associated bloodstream infections and surgical-site infections (28,29). Given the lack of standardization in making a diagnosis of pneumonia, it is unclear how much progress has been made toward a reduction in pneumonia-related HAIs (30). Our results leave little doubt that pneumonia is the HAI with the greatest economic impact in cardiac surgery patients. The rigorous definition of HAIs in general is crucial for valid comparisons across hospitals and over time. All major HAIs in our study were identified using CDC definitions and adjudicated by an independent committee of infectious-diseases experts. However, this level of monitoring and adjudication does not always reflect clinical decision making. For example, a patient who was treated for pneumonia may not have met the diagnostic criteria for pneumonia. Therefore, by using stricter criteria in this study than the criteria used for making clinical decisions, the economic burden of HAIs in this study may be a conservative estimate.
Our results show that the increases in LOS and cost associated with major HAIs during the index hospitalization remain substantial. Our findings are consistent with previous estimates of cost and LOS reported for postoperative infection in coronary artery bypass graft patients and with the results of a recent meta-analysis (5,15,31). However, they are higher than the estimated upper bound of attributable hospital cost due to major HAIs in the U.S. inpatient population ($25,903 per major HAI [year-2007 dollars]) (32).
Estimates of the incremental use of resources associated with HAIs vary significantly in published reports (33,34). Such variations are partly a result of the different settings and patient populations from which data were collected and partly a result of the different economic models used for each study. Infections acquired in surgical settings, for example, may have, on average, a greater impact on resource utilization than do those acquired by medically managed patients (35). Moreover, estimates from single-center studies will reflect the local case mix and practice of the center. The choice of statistical methodology can also lead to significant variation in the estimates of HAI-attributable cost and LOS (36–38). Regression analyses, such as GLM, are commonly the preferred method of accounting for the heterogeneity of a patient population. Matching strategies, which control for confounders in the design stage, rather than in the analysis stage, present a tradeoff that is related to the number of variables used for matching data from infected patients with those from noninfected controls (39). Too few variables may be insufficient to account for important variation (bias from omitted variables), whereas more variables may reduce the number of patients who can be compared in the cohort (selection bias). Regression models permit avoiding the selection bias that results from the exclusion of unmatched cases and controls while adjusting for a high number of confounders (36,38,39).
Readmissions related to HAI are frequent and more expensive than are non–HAI-related readmissions, costing, on average, 2.7-fold more. As a result, the cost of HAI-related readmissions represents about one-third of the total cost attributable to HAIs. Although there is evidence that higher readmission rates may not be associated with worse outcomes (and could even be associated with lower mortality), readmissions have been recently targeted as an indicator of presumed low-quality care that is associated with high costs (40). In particular, the Centers for Medicare and Medicaid Services now impose financial penalties on hospitals that have higher-than-expected 30-day readmission rates (risk standardized) after treatment for acute myocardial infarction, congestive heart failure, and pneumonia (41). Beginning in fiscal year 2017, such penalties are expected to include readmissions for patients undergoing coronary artery bypass graft surgery (42). The expansion of these policies will increase the importance of HAI prevention in cardiac surgery patients (43,44).
First, we captured only inpatient costs. The economic burden on patients and their families or other informal caregivers, as well as treatments provided in the outpatient setting, are not included. Second, cost data were available only for readmissions to the same hospital, which represented 64% of all readmissions. It might be possible that the cost of readmissions to nonindex hospitals was systematically different from those of readmissions to the same hospital, especially if the reason for change was dictated by emergency, or if the readmitting hospital lacked a cardiac surgery program, which could potentially lead to costlier and inferior care. However, the estimates of LOS, an important surrogate of resource utilization and reported in this study, also include out-of-network hospitalizations. Third, we did not examine the relationship between HAI-associated costs and specific types of cardiac procedures. However, the incremental cost and LOS data generated for a cohort of patients that excludes transplant and VAD recipients were not significantly different from those generated using the entire cohort. This finding suggests that our estimates of HAI-associated cost and LOS may be broadly applicable to cardiac surgery patients. Finally, our choice of model to adjust for measured confounders (GLM) does not account for the time-dependent nature of HAIs (45). However, GLM does allow for the adjustment of those baseline variables that had been measured in the observational study, which is an advantage over multistate models that adjust exclusively for time dependency (36,39).
This study identified the types of major HAIs in cardiac surgery patients, the effects of these HAIs on readmissions, and the utilization of specific hospital resources. It supports the widely held belief that reducing the enormous infection-related tolls of mortality and morbidity is not only a clinical imperative, but, especially in the current economic environment, an economic necessity. Every major HAI, on average, increases LOS by 2 weeks, adding nearly $38,000 to the cost of the index hospitalization. Moreover, readmissions associated with major infections have an LOS twice as long as those of other readmissions (11.5 vs. 6.0 days), and cost nearly threefold as much ($33,512 vs. $12,742). This study, therefore, substantiates the economic argument for preventive interventions and specifies the possible economic returns from such strategies. This information may help to drive quality-improvement initiatives to reduce HAIs and ultimately improve patient outcomes.
COMPETENCY IN PATIENT CARE: Knowledge of the costs of health care–associated infections can align clinical and economic decisions to facilitate efficient, cost-effective health care.
TRANSLATIONAL OUTLOOK: Although considerable cost savings can accrue from the prevention of health care–associated infections, additional research is necessary to evaluate the cost-effectiveness of various strategies of infection control.
This research was funded by the National Heart, Lung and Blood Institute; the National Institute of Neurological Diseases and Stroke (7U01 HL088942); the Canadian Institutes of Health Research; and the Institute for Health Technology Studies (InHealth). Dr. Alexander has received research support from Bristol-Myers Squibb, Boehringer-Ingelheim, CSL Behring, Tenax Therapeutics, Regado Biosciences, and Vivus Pharmaceuticals; and consulting fees from Bristol-Myers Squibb, Portola, and Regado Biosciences. Dr. Gelijns has served on the research advisory council of the now-defunct InHealth, a nonprofit organization that sponsored research on medical devices. All other authors report that they have no relationships relevant to the contents of this paper to disclose.
- Abbreviations and Acronyms
- Centers for Disease Control and Prevention
- Cardiothoracic Surgical Trials Network
- generalized linear model
- health care-associated infections
- International Classification of Diseases-Ninth Revision
- intensive care unit
- length of stay
- University HealthSystem Consortium
- ventricular assist device
- Received April 23, 2014.
- Revision received September 18, 2014.
- Accepted September 19, 2014.
- American College of Cardiology Foundation
- Ridderstolpe L.,
- Gill H.,
- Granfeldt H.,
- Ahlfeldt H.,
- Rutberg H.
- (2014) National and State Healthcare-Associated Infections Progress Report 2012 (Centers for Disease Control and Prevention, Atlanta, GA).
- Gelijns A.C.,
- Moskowitz A.J.,
- Acker M.A.,
- et al.
- Dixon R.E.
- Aronsky D.,
- Haug P.J.,
- Lagor C.,
- Dean N.C.
- Fu A.Z.,
- Qiu Y.,
- Radican L.,
- Wells B.J.
- Klevens R.M.,
- Edwards J.R.,
- Richards C.L. Jr..,
- et al.
- Whitehouse J.D.,
- Friedman N.D.,
- Kirkland K.B.,
- Richardson W.J.,
- Sexton D.J.
- Eagle K.A.,
- Guyton R.A.,
- Davidoff R.,
- et al.
- Patel M.R.,
- Dehmer G.J.,
- Hirshfeld J.W.,
- Smith P.K.,
- Spertus J.A.
- ↵Pronovost PJ, Marsteller JA, Goeschel CA. Preventing bloodstream infections: a measurable national success story in quality improvement. Health Aff (Millwood) 2011;30:628–34.
- Berenguer C.M.,
- Ochsner M.G. Jr..,
- Lord S.A.,
- Senkowski C.K.
- ↵Scott DR. The direct medical costs of healthcare associated infections in U.S. hospitals and the benefits of prevention. Atlanta, GA: Centers for Disease Control and Prevention. Available at: www.cdc.gov/HAI/pdfs/hai/Scott_CostPaper.pdf. Accessed December 2014.
- Roberts R.R.,
- Hota B.,
- Ahmad I.,
- et al.