American Journal of Respiratory and Critical Care Medicine

Fifty-seven patients developed an episode of catheter-related infection (CRI) in the bloodstream during their stay in the intensive care unit (cases) and were prospectively observed to establish the attributable mortality, increase in length of stay, and excess costs. Costs were estimated by multiplying the number of excess days of stay by the reimbursement provided. The outcomes for these cases were compared with those for matched control subjects without CRI. Eight cases were excluded as no control was found. Of the 49 cases, 31 were coagulase-negative staphylococci (CNS). The level of severity was similar for both groups (APACHE II 15.5 ± 7.2 versus 15.2 ± 7.3). There were no significant differences (p > 0.20) in the mortality observed in the hospital for the cases (22.4%, 95% confidence interval [CI] 0.3% to 34.9%) and the control subjects (34.7%, 95% CI 21.2% to 40.1%). Among the survivors, the hospital stay was increased by 19.6 d (95% CI − 1.1; 40.4). This represents an added cost of 3,124 Euros per episode of CRI among the survivors. In conclusion, our cohort study failed to show a difference in attributable mortality due to CRI in intensive care unit patients. Nevertheless, these infections lead to an increase in hospital stay of approximately 20 d. Each episode of CRI represents an additional cost of more than 3,000 Euros.

Bloodstream infections are the first or second cause of nosocomial infection in the intensive care unit (ICU) (1, 2). They represent between 30 and 40% of all nosocomial infections (3). In spite of the fact that a high percentage of bloodstream infections (20–40%) are of unknown origin, most have an identifiable source (50–80%) (2, 4). The most frequent origins of bloodstream infections in the ICU are venous catheters, respiratory, intraabdominal, and surgical wound infections, taking into account the differences between each hospital and the different pathologies that prevail in each ICU (1).

The catheters infected with the greatest frequency are the central intravenous catheters (5). The microorganisms most frequently involved are gram positive (Staphylococcus epidermidis and Staphylococcus aureus), fungi (Candida spp.), and gram negative (Pseudomonas aeruginosa) (3).

During the Consensus Meeting held by the Spanish Association of Intensive Care Medicine and Coronary Units (SEMIUC) (6), it was determined that catheter-related infections (CRI) show a wide range of degrees of severity, including suspected infections that require removal of the catheter, as well as complicated bloodstream infections with endocarditis or metastastic foci that demand surgical treatment, as well as an extended stay. Furthermore, Pittet and coworkers (2) have demonstrated that overall 35% of excess mortality and a significant increase in hospital stay can be attributed directly to nosocomial bacteremia, intravenous lines being responsible for at least 20 (23%) of the 86 episodes of bloodstream infection. Recently, the Sepsis Data study detected that 13.7% of the infections are catheter-related, and related mortality accounts for 9.2% of patients with CRI (7).

The estimated cost of nosocomial infection is based on a deduction of the charges incurred due to the increase in the length of the hospital stay. Although this calculation underestimates the overall expenses (8), there is a tendency to assume that the cost per day of the stay is the cost that primarily contributes to the expense attributed to the infection (9). Because there is no other more accurate analytic accounting system, increase in hospital stay is one of the parameters that best reflects the cost attributed to nosocomial infection (10).

There is medical literature on the costs of nosocomial infections in general and in the ICU (2, 11-13), but there is little information that refers to the costs of CRI for patients in intensive care. The hypothesis of the study was that CRI do not significantly increase mortality but they do contribute to increasing the hospital stay, with the subsequent economic repercussions that this implies. The main objectives were determining the consequences of CRI acquired in ICUs in terms of attributable mortality, increase in length of stay, and excess costs.


The study was performed in Hospital de Sabadell (Barcelona, Spain), which has 650 beds. A total of 11 beds are in the ICUs, where medical as well as surgical patients are admitted.


A retrospective, observational, cohort (1 case:1 control) study was performed with variables collected prospectively.

Subjects of Study

All those patients who showed CRI during the period between January 1992 and January 1999 and were admitted for more than 48 h were included in the study. The patients who developed CRI during their stay in the ICU were defined as cases and those who did not show CRI were considered potential control subjects.


Case. The patients who showed positive blood cultures 48 h subsequent to admission and positive catheter culture for the same microorganism were defined as patients with CRI. For patients with coagulase-negative staphylococci (CNS), identical strains (compared by antibiogram) of microorganisms isolated from two or more positive blood cultures were required (14). Those patients who developed CRI from catheters inserted prior to admission to the ICU, as well as episodes of relapse in those patients who showed more than one CRI, were excluded.

Control. A control patient was identified for each case during the period of study, controlling the level of severity (15) (APACHE II) at admission (± 2 points), underlying disease (same diagnostic category) within the Knaus classification (APACHE II), and age (± 5 yr). Furthermore, each control subject was required to have a stay equivalent to or greater than the number of days the case had been admitted when CRI developed.

In the definition of CRI, the same microorganism (species, antibiogram) of a semiquantitative culture (Maki's technique) (16) or quantitative culture (Cleri's technique) (14) from a segment of the intravenous catheter and blood culture (obtained by direct venopuncture) was isolated for a patient with clinical symptoms of infection and no other apparent focus of infection.

The mortality observed in the cases and control subjects was determined. The difference in the mortality observed in both groups was calculated. The attributable mortality was defined as the excess mortality caused by CRI. The estimated mortality was calculated based on the mortality predicted according to the APACHE II (15) system.

The hospital stay of the survivors was calculated for the cases as well as the control subjects. This calculation was performed separately from the detailed calculations, as it was considered that the cost and stay would vary depending on the condition of the patients. The excess cost attributed to CRI for the survivors was considered to be the difference between the total cost of the cases and the control subjects.

The cost calculated was based on the difference between the cases and control subjects in terms of the hospital stay: The number of days the length of stay was increased, multiplied by the cost of the stay/ 159.3 Euros per day (reimbursement provided by the Catalan Health Service for each day of hospitalization in 1996), was considered to be the economic cost of CRI.

Statistical Analysis

The stays were compared using Wilcoxon's nonparametric test (17). Mortality was compared by McNemar's test. The confidence intervals of 95% (95% CI) were calculated using the standard error derived from Taylor's method (18).

During the period between January 1992 and December 1998, of a total of 2,000 patients admitted for more than 48 h, 77 (3.85%) patients with CRI were identified. Ten patients were excluded because they showed CRI prior to admittance to the ICU; 9 due to recurrences and 1 because the patient was still admitted at the time of the study. In reality, there were 57 patients with CRI, which represents an incidence of 2.85 episodes per 100 admissions to the ICU. Eight patients were excluded from the study because no suitable control subject was identified (three patients due to underlying disease and five due to fewer days of catheterization). Two of these patients died in the ICU, whereas for the survivors the hospital stay was equivalent to 47.3 ± 19.3 d.

The clinical characteristics of the cases and control subjects are shown in Table 1. The degree of severity at admission according to APACHE II was similar for both groups (15.5 ± 7.2 versus 15.2 ± 7.3). Based on the degree of matching of the three main variables (APACHE, diagnostic category, and age), a coincidence of 86% (127/147) was observed. In brief, all cases (49/49, 100%) were matched for severity of illness at admission (APACHE II ± 2 points) and primary diagnosis based on Knaus' classification. The underlying disease of these patients is indicated in Table 2. Some discrepancies were due to age. Among cases and control subjects not successfully matched for age according to our criterion (± 5 yr), the difference never exceeded 10 years among matched pairs younger than 50 yr or 15 yr among matched pairs older than 50 yr. The cases had a mean age of 60.3 yr (median, 67 yr) and the controls a mean age of 58.4 yr (median, 67 yr). The total number of variables for matching (Table 1) was 882; we were able to fulfill 735 for a success rate of 83.3%. All control patients met the criterion of length of stay greater than or equal to the interval from admission to bloodstream infection in cases.


CharacteristicsValue of Characteristics Indicated
Cases (n = 49)Control Subjects (n = 49)p Value
Average age, yr ± SD60.3 ± 18.558.4 ± 20.5> 0.20
No. of men /women35/1433/16> 0.20
No. (%) with previous trauma4 (8.1)4 (8.1)> 0.20
Surgical history22 (44.8)21 (42.8)> 0.20
Corticoids2 (4)1 (2)> 0.20
Renal failure4 (8.1)3 (6.1)> 0.20
COPD7 (14.2)6 (12.2)> 0.20
Cardiopathy10 (20.4)14 (28.5)> 0.20
Coma6 (12.2)6 (12.2)> 0.20
MV38 (77.5)38 (77.5)> 0.20
ARDS8 (16.3)9 (18.3)> 0.20
Solid neoplasia4 (8.1)4 (8.1)> 0.20
Shock (nonseptic)3 (6.1)5 (10.2)> 0.20
APACHE II (at admission)15.5 ± 7.215.2 ± 7.3> 0.20

Definition of abbreviations: ARDS = adult respiratory distress syndrome; COPD = chronic obstructive pulmonary disease; CRI = catheter-related infection; MV = mechanical ventilation.


Diagnostic CategoryCases (n = 49)Control Subjects*(n = 49)
Nonoperated patients
 Cardiovascular failure due to sepsis8 (4%)8 (4%)
 Multiple traumas6 (3%)6 (3%)
 Cardiovascular failure due to coronary disease5 (2.5%)5 (2.5%)
 Cardiovascular failure due to CRA4 (2%)4 (2%)
 Respiratory failure due to aspiration/toxics3 (1.5%)3 (1.5%)
 Respiratory failure due to COPD4 (2%)4 (2%)
 Respiratory failure due to ARDS1 (0.5%)1 (0.5%)
 Respiratory failure due to infection3 (1.5%)3 (1.5%)
 Neurological4 (2%)4 (2%)
 Cardiovascular failure due to CHF1 (0.5%)1 (0.5%)
 Gastrointestinal disease1 (0.5%)1 (0.5%)
 Cranial trauma1 (0.5%)1 (0.5%)
 Cardiovascular failure due to alteration of rhythm1 (0.5%)1 (0.5%)
Operated patients
 Gastrointestinal surgery due to neoplasia2 (1%)2 (1%)
 Intestinal obstruction/perforation3 (1.5%)3 (1.5%)
 Multiple trauma2 (1%)2 (1%)

Definition of abbreviations: ARDS = adult respiratory distress syndrome; CHF = congestive heart failure; CRA = cardiorespiratory arrest; COPD = chronic obstructive pulmonary disease.

*p value > 0.20.

The mean interval between admission to the ICU and identification of CRI was 14.4 ± 14.8 d (median of 8 d). Thirty-one (63.2%) CNS, 5 (10.2%) Staphylococcus aureus, 5 (10.2%) Pseudomonas aeruginosa, 2 (4.0%) Klebsiella sp., 1 (2.0%) Citrobacter sp., 1 (2.0%) Enterococcus sp., 1 (2.0%) Serratia sp., 1 (2.0%) Candida albicans, 1 (2.0%) Escherichia coli, and 1 (2.0%) Streptococcus viridans were isolated. Among the patients with positive blood cultures for coagulase-negative staphylococci mortality was 19.3%, whereas the excess hospitalization for survivors with CRI caused by CNS, in comparison with their control subjects (the matched survivors without CRI), was 31.4 ± 52.0 d (p = 0.02). For the other etiologies, mortality was 27.8% and excess hospitalization was 0.1 ± 39.4 d, with no evidence of statistical differences (p > 0.20) in comparison with the control subjects.

The mortality observed and estimated is shown in Table 3. Among the survivors, the mean hospital stay for the cases was 63.1 ± 51.0 d (median of 18 d), whereas for the controls this figure was 43.4 ± 24.7 d (median of 19 d). This represents an excess hospitalization of 19.6 ± 49.2 d (95% CI −1.1; 40.4), and an additional cost of 3,124 Euros for each survivor of a CRI episode.


Cases (n = 49)Controls (n = 49)p Value
Observed mortality in
  hospital, n (%)11 (22.4%)17 (34.7%)
 95% CI 10.3% to 30.9%21.2% to 40.1%> 0.20
Observed mortality in
  ICU, n (%)9 (18.4%)14 (28.6%)> 0.20
 95% CI 6.9% to 30.1%14.8% to 42.3%
Estimated mortality,* %23.7% ± 10.524.3 ± 10.1> 0.20

Definition of abbreviations: CI = confidence interval; ICU = intensive care unit.

*  The estimated mortality was derived based on the mortality predicted according to the APACHE II (13) system.

There is significant controversy as to whether the critically ill patients who are infected die as a result of CRI, or whether the infection is a terminal event that occurs prior to imminent death. This is an essential issue, as it justifies the efforts made to prevent the development of such infections. Unfortunately, because a large number of patients must be evaluated, these studies are difficult to perform. Therefore, there is limited information available on this subject, and it was a controversial topic at the Consensus Meeting on catheter-related infection organized by the SEMIUC (7). In our cohort study, the differences between cases and control subjects in terms of mean mortality was not significant. This observation contrasts with the observation made by Pittet and coworkers (2) on the group of infections as a whole of surgical ICU inpatients. Such findings may also have been influenced by the fact that most of the episodes were due to CNS, which, in our institution, were treated regularly with vancomycin, and in all cases the catheter was removed. The possibility that maintaining the catheter or introducing nonsystematic treatment with antibiotics might yield different results cannot be ruled out. In addition, S. aureus or Candida infections are associated with a much higher mortality than CNS. Because the majority of patients studied had coagulase-negative staphylococcal bloodstream infections, a failure to show a difference in mortality in this cohort may merely reflect the preponderance of a less pathogenic organisms as the cause of most bloodstream infections in this study. In any case, a large multicentric study on nosocomial bacteremia reported from Spanish ICUs (4) suggests that, unlike other origins (particularly intraabdominal or respiratory infection), intravenous catheters were associated with a less significant impact on mortality.

On the other hand, among the survivors, we have found an increase in the hospital stay of approximately 20 d, similar to the values found by Pittet and coworkers (2) in their study on bloodstream infections in general. Other researchers have found similar results for candidemias or bacteremias developed in the hospital (19-21). Martin and coworkers (22) have found that bacteremia related to CNS increased the hospital stay by 8.5 d. They suggest that although it was unable to cause attributable mortality, it contributed to increasing morbidity. Two recent studies in critically ill patients also support these findings (23, 24). The information provided by the current study offers an interesting means to estimate the direct costs of excess hospitalization attributed to CRI in the patients in the ICU in our institution. Based on the hospitalization costs in other centers, it is relatively easy to extrapolate these data to other institutions. Although CRI affects less than 3% of the patients admitted to the ICU, taking into account the mortality observed in these patients, it costs approximately 7,813 Euros per 100 patients admitted. Therefore, the investment in preventive measures is justified. Since nearly 50% of our casemix refers to coronary patients, who rarely develop nosocomial infections, the real costs on a patient mix of noncoronary intubated patients could be twice our estimate.

In the interpretation of the results, the degree of coincidence in the matching between cases and control subjects is critical. We have introduced an adjustment based on the level of severity at admission, the main diagnostic category, and the age, as these are the most important variables as prognostic indicators. The effect of matching our population at the time of admission to the ICU rather than at the time the bloodstream infection was recognized may have affected our results (24). Of the 882 variables analyzed in our cohort, 735 (83.3%) were successfully matched. Furthermore, all of the control subjects were exposed to risk (days of catheterization) during a period equivalent to the minimum number of days the cases were exposed. This condition is fulfilled only partially in other studies (2, 15), and was the main reason for exclusion of the control subjects. Nevertheless, among the cases excluded due to the fact that no control subjects were found, no significant differences were observed as regards mortality and stay of the survivors. This suggests that the inclusion of these cases in the cohort studied would not modify our findings.

Other limitations are inherent to retrospective studies, although they were minimized by the fact that the data base was constructed prospectively. The main limitation was that the number of cases is rather limited, in spite of the extended period in which the cases were compiled. This is deduced from the wide range of values observed in the CIs. The most important implication of the very wide CI is that our estimates of exactly how many days and how much extra money are rather uncertain. In addition, that a 14% higher death rate in control subjects was not significant also illustrates the fact that this is a rather small sample size unable to detect such a difference as significant. Therefore, it would be desirable to perform multicentric studies that verify these findings, particularly to evaluate specific etiologies. Unfortunately, we do not know of a study of this type referring to CRI published in the literature. Due to the number of patients who must be considered, it is not likely that studies of this type will be available in the short term.

Second, our adjustment was based on the severity of illness at admission, but the degree of severity of illness prior to onset of nosocomial infection, as demonstrated in a recently published study by Soufir and coworkers (23), may have drifted apart for the cases and control subjects. A differing level of severity of illness just before onset of bloodstream infection in cases could explain the higher expected mortality in the patients selected as control subjects. Finally, it must be pointed out that a minority of our patients were receiving parenteral nutrition by the catheter considered. Therefore, these results cannot be generalized to catheters used exclusively for parenteral nutrition or for other specific purposes (for example, for arterial purposes or hemodialysis). Neither can it be generalized for hospital patients outside of the ICU or with devices that are difficult to replace (Portacath).

In summary, our data suggest that the prevention of CRI may not decrease the mortality of ICU inpatients. Rather, among the survivors there is a significant increase in hospital stay, which occurs regardless of the reason for admission, age, or level of severity. These data can be used to evaluate the potential economic benefits foreseen due to the application of new techniques in the prevention of intravascular CRI.

Supported in part by grants from CIRIT (SGR97/443) and FIS (99/038).

1. Vallés, J. 1997. Bacteriemias en Cuidados Intensivos. Enferm. Infecc. Microbiol. Clin. 15(Suppl. 3):8–13.
2. Pittet D., Tarara D., Wenzel R. P.Nosocomial bloodstream infection in critically ill patient. Excess length stay, extra cost, and attributable mortality. J.A.M.A.271199415981601
3. Wenzel R. P., Thompson R. L., Landry S. M., Russel B. S., Miller P. J., Ponce de León S., Miller G. B.Hospital acquired infections in intensive care unit patients: an overview with emphasis on epidemiologics. Infect. Control41983371375
4. Vallés J., León C., Álvarez F.Spanish Collaborative Group for Infections in ICUsNosocomial bacteremia in critically ill patients: a multicenter study evaluating epidemiology and prognosis. Clin. Infect. Dis.241997387395
5. Raad I., Darouiche R.Prevention of infections associated with intravascular devices. Crit. Care Med.21996361365
6. Olaechea, P. M., T. Munoz, S. Gallardo, and I. Areitio. Repercusiones económicas de las infecciones por catéter. monografı́a médica: infección por Catéter en UCI SEMIUC. Fernández Ed. Madrid 129–140.
7. Navas E.Bacteriemia asociada a catéter. Rev. Clin. Esp.19719974144
8. Haley R. W., Shaberg D. R., Crossley K. B., Von Allmen S. D., McGoman J. E.Extra charges and prolongation stay attributable to nosocomial infections: a prospective inter hospital comparison. Am. J. Med.7019815158
9. Wakefield D. S., Helms Ch. M., Massanari R. M., Mor M., Pfaller M.Cost of nosocomial infection: relative contributions of laboratory, antibiotic and per diem cost in serious Staphylococcus aureus infections. J. Infect. Control161988185192
10. Peña C., Pujol M., Pallarés R., Corbella X., Vidal T., Tortras N., Ariza J., Gudiol F.Estimación del coste atribuible a la infección nosocomial: prolongación de la estancia hospitalaria y cálculo de los costes alternativos. Med. Clin. (Barc.)1061996441444
11. De Clerq H., De Decker G. H., Alexander J. P., Huyghens L.Cost evaluation of infections in intensive care. Acta Anaesthesiol. Belg.341983179189
12. Diaz Molina, C., M. Garcı́a Martı́n, A. Bueno Cavanillas, A. Lopez Luque, M. Delgado Rodriguez, and R. Gálvez VargasEstimación del coste de la infección nosocomial en una unidad de medicina intensiva. Med. Clin. (Barc.)1001993329332
13. Herman B., Wolff M. A., Pean Y., Gertner J., Millet H.Risques infectieux liés au catheterisme veineux et essai déstimation de leur cout financier. Agresologie261985197202
14. Rello J., Quintana E., Mirelis B., Gurgui M., Net A., Prats G.Polymicrobial bacteremia in critically ill patients. Intens. Care Med.1919932225
15. Rello J., Jubert P., Vallés J., Artigas A., Rué M., Niederman M.Evaluation of outcome for intubated patients with pneumonia due to Pseudomonas aeruginosa. Clin. Infect. Dis.231996973978
16. Maki D. G., Weise C. E., Sarafin H. W.A semiquantitative culture method for identifying intravenous catheter-related infection. N. Engl. J. Med.296197713051309
17. Conover, W. J. 1980. Practical Nonparametric Statistics, 2nd ed. John Wiley & Sons, New York. 363–364.
18. Bishop, Y. M. M., S. E. Feinberg, and P. W. Holland. 1975. Discrete Multivariate Analyses: Theory and Practice. Massachusets Institute of Technology Press, Cambridge. 492–497.
19. Spengler R. F., Greenough W. B.Hospital costs and mortality attributed to nosocomial bacteremias. J.A.M.A.240197824552458
20. Rose R., Hunting K. J., Towsend T. R., Wenzel R. P.Morbidity/mortality and economics of hospital acquired bloodstream infections: a controlled study. South. Med. J.70197712671269
21. Wey S. B., Motomi M., Pfaller M. A., Woolson R. F., Wenzel R. P.Hospital-acquired candidemia: the attributable mortality and excess length of stay. Arch. Intern. Med.148198826422645
22. Martin M. A., Pfaller M. A., Wenzel R. P.Coagulase-negative staphylococcal bacteremia: mortality and hospital stay. Ann. Intern. Med.1101989916
23. Soufir L., Timsit J.-F., Mahe C., Carlet J., Regnier B., Chevret S.Attributable morbidity and mortality of catheter-related septicemia in critically ill patients: a matched, risk-adjusted cohort study. Infect. Control Hosp. Epidemiol.201999396401
24. DiGiovine B., Chenoweth C., Watts C. H., Higgins M.The attributable mortality and costs of primary nosocomial bloodstream infections in the intensive care unit. Am. J. Respir. Crit. Care Med.1601999976981
Correspondence and requests for reprints should be addressed to Jordi Rello, M.D., Ph.D., Critical Care Department, Hospital Universitari Joan XXIII, Carrer Dr. Mallafre Guasch, 4, 43007 Tarragona, Spain. E-mail:


No related items
American Journal of Respiratory and Critical Care Medicine

Click to see any corrections or updates and to confirm this is the authentic version of record