Objective. To determine the attributable cost of nosocomial primary bloodstream infections (BSIs) in PICU patients.
Methods. A prospective cohort study was conducted of the PICU of the St Louis Children's Hospital, a 235-bed academic tertiary care center. All patients who were admitted to the PICU were included unless they met the following exclusion criteria: age >18 years, death within 24 hours of PICU admission, admission to the NICU service. Total and direct medical costs of PICU and hospital stay for patients with and without nosocomial primary BSI were measured.
Results. Fifty-seven children developed 65 episodes of primary BSIs during their PICU stay. The rate of BSI in this population was 13.8 per 1000 central venous catheter days. In multiple linear regression analysis, severity of illness as measured by the admission Pediatric Risk of Mortality Score III, congenital heart disease, underlying lung disease, ventilator days, transplant (solid organ and bone marrow), and nosocomial primary BSI were independent predictors of PICU direct costs. The direct cost of PICU admission for patients with nosocomial primary BSI was $45 615 and for the patients without primary BSI was $6396.
Conclusions. After controlling for age, severity of illness, underlying disease, and ventilator days, we found that the direct cost of PICU admission attributable to nosocomial primary BSI was $39 219. The prevention of these infections through specific interventions is likely to be cost-effective.
Nosocomial primary bloodstream infection (BSI) is the most common hospital-acquired infection in PICU patients, causing nearly one third of all nosocomial infections in this patient population.1 Studies of the attributable cost of nosocomial primary BSI in adult ICU patients estimate that these infections cost up to an additional $40 000.2–6 The costs that are attributable to primary BSI may be very different in PICU patients, as they are a heterogeneous group of patients who have a variety of ages and underlying diseases and undergo different surgeries and procedures from adult ICU patients. In addition, the overall mortality among PICU patients is much lower than adult ICU patients, which may affect costs.
Few studies of the cost of nosocomial primary BSIs have been performed in PICU patients. Dominguez et al7 performed a case-control study nested within a prospective cohort and found that the average total cost for an infectious complication in PICU patients was $50 362. The authors matched patients on diagnostic category, length of stay (±1 day), and Pediatric Risk of Mortality Score III (PRISM; ±10 points). This study was limited by a small number of cases (only 23 of the 30 patients could be matched); thus, 1 or 2 patients with expensive hospitalizations could potentially result in an overestimation of the average cost. The calculation of cost was based on charge, using cost-to-charge ratios, which includes a fixed component of indirect costs that are a less accurate estimate of attributable costs.8 Finally, no analysis of the costs of different types of nosocomial infection (BSI vs ventilator-associated pneumonia vs surgical site infection vs central nervous system infection) was performed.
A second matched case-control study, published by Slonim et al,9 examined the attributable costs of nosocomial primary BSI in PICU patients. Cases were matched to control subjects by age, severity of illness, primary diagnosis, and admission date. These investigators found that nosocomial BSIs cost an average of $46 133 in total operating costs.9 As in the study by Dominguez et al, the study was limited by estimates of costs that were obtained using charges and cost-to-charge ratios, as well as a relatively small number of cases (n = 38). We performed a prospective cohort study to determine the attributable costs of nosocomial primary BSI in PICU patients, using a multiple linear regression model to control for other potential predictors of costs.
St Louis Children's Hospital is a 235-bed academic tertiary care center affiliated with Washington University School of Medicine. It serves a 300-mile radius referral base in southwestern Illinois and southeastern Missouri. The St Louis Children's Hospital PICU was a 22-bed PICU at the initiation of the study, increasing to 28 beds in April 2000, with ∼1400 admissions per year. The maximum patient-to-nurse staffing ratio is 2:1.
This was a prospective cohort study of all patients who were admitted to the St Louis Children's Hospital PICU from September 1, 1999, to May 31, 2000. All patients who were admitted to the PICU were eligible for study admission unless they met the following exclusion criteria: (1) age >18 years, (2) death within 24 hours of PICU admission, or (3) a NICU patient on extracorporeal membrane oxygenation occupying PICU bed space. Institutional Review Board approval was obtained from the Washington University School of Medicine; a waiver of written informed consent was requested and granted because of the observational nature of the study.
The Centers for Disease Control and Prevention and the National Nosocomial Infection Surveillance System definitions were used for nosocomial primary BSI.10 Specifically, nosocomial primary BSI was defined as a recognized pathogen isolated from blood culture after 48 hours of PICU admission. Coagulase-negative Staphylococcal species was reported as a cause of BSI only when 2 or more blood cultures drawn on different occasions were positive for coagulase-negative Staphylococcal species. Staphylococcal strains were not speciated.
Data were collected on additional potential clinical predictors of cost, including severity of illness, underlying lung disease, congenital heart disease, transplantation, ventilator days, and age. Admission severity of illness was measured using the PRISM, a weighted score of 17 physiologic parameters obtained during the first 24 hours of admission, developed by Pollack et al.11 Underlying lung disease was defined as a history of chronic pulmonary conditions as documented in the resident or attending hospital admission note, including asthma, cystic fibrosis, and bronchopulmonary dysplasia, and/or the use of home mechanical ventilation or supplemental oxygen and/or abnormal pulmonary function tests. Congenital heart disease was defined as a patient who was born with cardiac anatomy other than the following: 4 chambers, the absence of an intracardiac shunt (excluding patent foramen ovale), right ventricle drained by the pulmonary artery and left ventricle drained by the aorta, normal valves, and normal coronary arteries. Transplant was defined as any type of solid organ (eg, lung, heart, liver, kidney) or bone marrow transplant. Ventilator days were measured from admission to the PICU until discharge. Age was defined as age at PICU admission, measured in days.
Costs were estimated from the perspective of the hospital as provider. Data were obtained from the hospital's cost accounting system (McKesson-HBOC, Atlanta, GA), which uses a ratio of cost-to-charges method to estimate costs and allocates overhead using a stepdown method. Total costs included direct costs, which are costs incurred in the provision of care to patients, and indirect costs, which includes costs for other support activities in the hospital not directly used in patient care. Separate analyses were performed for total costs and direct costs. In addition, both total costs and direct costs were stratified by department so that comparisons could be made for cost components of room and board, laboratory, supplies, pharmacy, radiology, and operating room. All costs were incurred in 1999 and 2000. Because of the small time window, no discounting was applied.
Data Collection and Measurement
Data were collected prospectively by 1 of the investigators (A.E.), who reviewed the patients' charts, bedside flow sheets, laboratory reports, and radiology reports. Data were double entered into an Access database (Microsoft Corp, Redmond, WA).
Data were analyzed using SPSS version 10.0 (SPSS, Chicago, IL). Univariate analysis was performed to determine associations between clinical characteristics and nosocomial primary BSIs, using a χ2 test to compare categorical variables and the Mann Whitney U test to compare continuous variables. Multiple linear regression was used to model the natural logarithm of PICU direct costs. The log transformation was necessary as the direct costs of PICU admission were not normally distributed. Variables that were believed to be clinically relevant or that a priori would be expected to predict costs were selected for inclusion in the model. The model was checked for first-degree interactions between significant variables. Diagnostic tests for collinearity were also performed. Outlying and influential cases were identified using standardized residuals, Cook's D test, Leverage values, and standardized changes in the regression coefficient. Linear regression models were run with and without extreme outliers. Mean and median costs for infected and uninfected patients were calculated for each area and compared for statistical significance using the Mann Whitney U test.
The attributable direct costs of BSI in the PICU were estimated as the difference in expected costs between patients with BSI and those without according to the following formula for a conditional expected mean: where C is the attributable cost of infection, βi (i = 1 … k) is an estimated regression coefficient with βk as the coefficient for BSI, and xi (i = 1 … k) are the independent variables in the regression, with xk as the binary indicator for bloodstream infection.
There were 911 PICU admissions between September 1, 1999, and May 31, 2000. Patient demographics are reported in Table 1. There was a white predominance (74%). Slightly more than half (58%) of the patients were male. Approximately half of the patient population was younger than 3 years. Congenital heart disease (29%), lung disease (25%), and genetic syndrome (18%) were common. Fewer than 5% of the patient population were admitted because of immunodeficiency, diabetes, renal disease, multiple trauma, or burn. Fifty percent of the patients had an arterial catheter, and 12% had multiple central venous catheters. Seventy percent were mechanically ventilated, and 30% had 1 or more transfusions during their stay in the PICU.
A comparison of the characteristics of patients with and without primary BSIs is summarized in Table 1. These data were previously published in Pediatrics. Only factors that occurred in >10% of patients are reported in Table 1. Age, gender, and ethnicity were not significantly different between patients with and without primary BSIs. Severity of illness at admission (PRISM III) was higher in patients with BSIs. A higher proportion of patients with BSI had received transplants before PICU admission and were treated with immunosuppressive agents, histamine type 2 receptor blockers, total parenteral nutrition, antimicrobial agents, and steroids during PICU admission. Patients with underlying medical conditions or invasive devices and those who underwent surgeries had an increased risk for developing BSIs in the univariate analysis.
Fifty-seven children developed 65 episodes of primary BSIs during their PICU stay. Five episodes of BSI were polymicrobial, and 7 patients had multiple BSIs. The rate of BSI in this population was 13.8 per 1000 central venous catheter days. The most common causative microorganisms are reported in Table 2. Gram-positive bacteria were responsible for most of these infections (55%). Coagulase negative Staphylococcus species (n = 28) were the leading cause of BSI, followed by Enterobacter cloacae (n = 8). Candida species were responsible for all 5 episodes of fungemia. There were insufficient numbers of cases caused by any 1 organism to determine relationships between organism type and patient characteristics. The mean time from PICU admission to BSI was 11.7 days (median: 10 days; range: 2–33 days).
In univariate analysis, the mean total costs of hospital admission were $80 291 for the infected patients and $15 663 for the uninfected patients (P = .001). The mean direct costs of PICU admission were $70 936 for the infected patients and $10 828 for the uninfected patients (P = .0001). The analysis of cost by area revealed that mean direct costs for room and board for patients with BSIs was $11 672 compared with $7385 for uninfected patients (P = .370; Table 3) and that mean total costs for room and board for patients with BSIs was $24 619 compared with $14 990 for uninfected patients (P = .357; Table 4). Direct laboratory costs were also higher for infected patients (mean: $2333 vs $2190; P = .974), but the difference was not statistically significant (Table 3).
In multiple linear regression analysis, severity of illness as measured by the admission PRISM, congenital heart disease, underlying lung disease, ventilator days, transplant, and nosocomial primary BSI were independent predictors of PICU direct costs (Table 5). Age was not a statistically significant predictor of PICU direct costs (Table 5). These predictors all remained statistically significant even after the extreme outliers, and influential cases (n = 7 outliers and influential cases, 6 of whom had BSI) were excluded from the multiple linear regression model.
Using the formula previously described, we calculated the direct costs of PICU admission for patients with nosocomial primary BSI as $45 615 and for the patients without primary BSI as $6396. The difference between the PICU direct costs for the infected and uninfected patients was $39 219.
We performed a prospective cohort study, using costs from the hospital accounting department and a multiple linear regression model to control for other potential predictors of PICU admission costs, to determine the attributable costs of nosocomial primary BSI in PICU patients. We found an excess of $39 219 in direct PICU costs among patients with nosocomial primary BSI after controlling for age, severity of illness, underlying disease, and ventilator days. Our findings indicate a higher attributable cost of nosocomial primary BSI than that found in the 2 other published studies, as we calculated direct costs of PICU admission rather than total PICU and/or hospitalization costs. We identified other predictors of PICU direct cost, including severity of illness, congenital heart disease, lung disease, transplant, and ventilator days. Congenital heart disease, lung disease, and transplant have not previously been described as predictors of cost in PICU patients.
There are several potential reasons for the differences in cost in our study compared with previous studies. First, we had a larger sample size and included more patients with nosocomial primary BSI (n = 57 vs 30 and 40 in the studies by Dominguez et al and Slonim et al, respectively)7,9 and were able to use cost data on all 57 patients, given our use of a multiple linear regression model to control for other potential confounders, rather than matching as in the 2 previous studies. Matching was not performed successfully on all of the cases in the previous studies, further decreasing the number of patients analyzed. A larger data set may permit more accurate sampling of the overall patient population and a more accurate estimate of the true attributable costs. A second difference between our study and the 2 previous studies is the methods. We used actual costs as opposed to charges and cost-to-charge ratios. Actual costs are a more accurate method than charges for estimating attributable costs. Third, our study was performed more recently (1999–2000) than the 2 previous studies (1993–19947 and 1996–19989), and process of care may have changed in the interim, increasing the costs. This is a less satisfying explanation for the differences between our findings and those of Slonim et al, given that there was only a 1-year interval between the conclusion of their study and the initiation of ours. Inflation may also have increased the costs. Finally, it is possible that the costs of PICU admission in our hospital are higher, given our transplant population and extracorporeal membrane oxygenation capability. Of note, transplant was included in our multivariate model and was a significant predictor of cost. However, we found the difference in direct PICU costs to be $39 000 even when controlling for the effect of organ transplantation on cost.
Studies using similar methods in other academic tertiary care hospitals are necessary to confirm our findings. Although we controlled for other putative predictors of cost, the possibility remains that there are other differences in the process of care provided by our PICU staff that are unique to our hospital and that increase cost.
Our study shows that among PICU patients with nosocomial primary BSI, the direct cost of PICU admission is significantly increased compared with other, noninfected patients with similar underlying diseases and numbers of ventilator days. The prevention of these infections through specific interventions is likely to be cost-effective.
This study was supported by National Institutes of Health Grant 1K23AI50750-01A1.
We thank Cherie Hill for assistance with database management and Cheryl Giubardo for assistance with obtaining cost data.
- Accepted August 16, 2004.
- Reprint requests to (A.M.E.) St Louis Children's Hospital, 1 Children's Place, Room 11W32, St Louis, MO 63110. E-mail:
No conflict of interest declared.
- ↵Richards MJ, Edwards JR, Culver DH, Gaynes RP, the National Nosocomial Infections Surveillance System. Nosocomial infections in the pediatric intensive care units in the United States. Pediatrics.1999;103(4) . Available at: www.pediatrics.org/cgi/content/full/104/4/e39
- Roberts RP, Scott RD, Cordell R, et al. The use of economic modeling to determine the hospital costs associated with nosocomial infections. Clin Infect Dis.2003;36 :1424– 1432
- ↵Haley RW, Shaberg DR, Von Allmen SD, McGowan JE. Estimating the extra charges and prolongation of hospitalization due to nosocomial infections: a comparison of methods. J Infect Dis.1980;141 :248– 257
- ↵Garner JS, Jarvis WR, Emori TG, Horan TC, Hughes JM. CDC definition for nosocomial infections. In: Olmsted RN, ed. Infection Control and Applied Epidemiology: Principle and Practice. St. Louis, MO: Mosby; 1996:A1– A20
- Yogaraj JS, Elward AM, Fraser VJ. Rate, risk factors and outcomes of nosocomial primary bloodstream infection in pediatric intensive care unit patients. Pediatrics.2002;110 :481– 485
- Copyright © 2005 by the American Academy of Pediatrics