Skip to main content

Development and validation of a multivariable prediction model of central venous catheter-tip colonization in a cohort of five randomized trials



The majority of central venous catheters (CVC) removed in the ICU are not colonized, including when a catheter-related infection (CRI) is suspected. We developed and validated a predictive score to reduce unnecessary CVC removal.


We conducted a retrospective cohort study from five multicenter randomized controlled trials with systematic catheter-tip culture of consecutive CVCs. Colonization was defined as growth of ≥103 colony-forming units per mL. Risk factors for colonization were identified in the training cohort (CATHEDIA and 3SITES trials; 3899 CVCs of which 575 (15%) were colonized) through multivariable analyses. After internal validation in 500 bootstrapped samples, the CVC-OUT score was computed by attaching points to the robust (> 50% of the bootstraps) risk factors. External validation was performed in the testing cohort (CLEAN, DRESSING2 and ELVIS trials; 6848 CVCs, of which 588 (9%) were colonized).


In the training cohort, obesity (1 point), diabetes (1 point), type of CVC (dialysis catheter, 1 point), anatomical insertion site (jugular, 4 points; femoral 5 points), rank of the catheter (second or subsequent, 1 point) and catheterization duration (≥ 5 days, 2 points) were significantly and independently associated with colonization . Area under the ROC curve (AUC) for the CVC-OUT score was 0.69, 95% confidence interval (CI) [0.67–0.72]. In the testing cohort, AUC for the CVC-OUT score was 0.60, 95% CI [0.58–0.62]. Among 1,469 CVCs removed for suspected CRI in the overall population, 1200 (82%) were not colonized. The negative predictive value (NPV) of a CVC-OUT score < 6 points was 94%, 95% CI [93%–95%].


The CVC-OUT score had a moderate ability to discriminate catheter-tip colonization, but the high NPV may contribute to reduce unnecessary CVCs removal. Preference of the subclavian site is the strongest and only modifiable risk factor that reduces the likelihood of catheter-tip colonization and consequently the risk of CRI.

Clinical Trials Registration: NCT00277888, NCT01479153, NCT01629550, NCT01189682, NCT00875069.


Central venous catheters (CVC), namely CVCs used to administer drugs and catheters used to deliver renal-replacement therapies (DC), are essential in the care of critically ill patients. In Europe in 2017, CVC use was 70.1 per 100 patient-days [1]. Up to two thirds of patients admitted in Intensive Care Units (ICU) are exposed to at least one CVC [2].

Unfortunately, their routine use is responsible for serious complications [3]. Among these, the most life threatening is catheter-related bloodstream infection (CRBSI). Although its incidence is decreasing over time with a current rate of 0.2 to 5/1000 catheters-days [1, 4,5,6], CRBSI is associated with mortality, prolonged ICU stays and higher hospitals costs [7,8,9]. While guidelines exist for the diagnosis and management, the strict definition of CRBSI relies on catheter-tip culture after catheter removal, or blood cultures, which may delay the diagnosis [10, 11]. When a CRBSI occurs, urgent removal of the CVC is the cornerstone of the management strategy. Obtaining positive blood cultures may be difficult under broad-spectrum antibiotics, which led to the catheter-related infection (CRI) definition (non-bacteremic clinical sepsis plus catheter-tip colonization or CRBSI). There is no consensus on how to define and act on suspected CRI, a very frequent situation in the ICU [10, 11].

Despite the importance of this topic, very few studies have investigated the optimal strategy in case of suspected CRI. Evidence for the safety of watchful waiting remains scarce. A randomized controlled trial (RCT) by Rijnders et al. [12] compared CVC removal left at the physician’s discretion (SOC) with watchful waiting amongst hemodynamically stable patients without proven bacteremia, no insertion site infection and no intravascular foreign body. The authors found no significant difference between the two groups in terms of CRBSI incidence and ICU mortality, but the sample size was small (n = 64). Of note, significantly fewer catheters were removed in the “watchful waiting” arm (62% reduction compared to the SOC arm, p < 0.01). Similar results were observed in a prospective observational study by Lorente et al. [13]. On the other hand, the absence of focus eradication (i.e., CVC removal) is clearly associated with mortality in the setting of CRBSI [14,15,16]. Consequently, the decision to remove the CVC is generally left at the physician’s judgement and experience. Of note, the fear of CRBSI on its own may trigger CVC replacement in a new anatomical site. However, systematic CVC replacement also exposes patients to potential harm [17]. These include mechanical complications when inserting a CVC in a new site. Moreover, the proportion of catheters that are not colonized after removal for suspected infection is significant in the series, up to 91% [18]. Developing tools to identify CVCs with a low likelihood of colonization could help reduce this proportion of premature CVC replacement. The objective of this study was to develop and validate a score capable of assessing the risk of CRI.

Material and methods

Study design

This report complies with the TRIPOD statement [19].

In this post-hoc analysis, we used the databases from five multicenter RCTs, for which data had been collected prospectively. We constructed the training cohort, used to develop the risk score, by merging 3SITES and CATHEDIA [3, 20]. Both studies investigated the impact of the anatomical insertion site on mechanical and intravascular catheter complications. The trials included in this training cohort were chosen by convenience because of the timing of availability of 3SITES and CATHEDIA databases. Subsequently, we used data from ELVIS, CLEAN and DRESSING2, which investigated different prevention strategies on the incidence of CRI [21,22,23], as testing cohort for external validation.

Study patients and catheters

Inclusion criteria slightly differed between the studies and are detailed in Additional File 1. All patients were adults admitted to ICU and requiring catheterization through a new venipuncture. Data for more than one catheter per patient could be recorded in the five studies (we therefore define the rank of the catheter as equal to one for the first catheter inserted during the ICU stay and > 1 for the second and subsequent catheters).

Regarding the current study, focus was made on CVCs and DCs, with exclusion of arterial catheters. Catheters with missing data concerning the colonization at removal were excluded.

Outcome and definitions

The principal outcome was colonization, defined as growth of ≥ 103 colony-forming units per milliliter from the catheter-tip culture according to Brun-Buisson quantitative technique [24]. Interventions were blinded to microbiologists who processed catheter cultures. CRI was defined as the association of colonization with local (pus or inflammation at the insertion site) or systemic signs such as fever (body temperature ≥ 38.5 °C) or hypothermia (body temperature ≤ 36.5 °C) in the absence of any other identified infection focus, corresponding to the CRI-1 and CRI-2 subgroups proposed by the European Centre for Disease prevention and Control (ECDC) [25]. Suspicion of CRI was based on clinical judgement, and the physicians caring for each patient independently made the decision to remove the CVC in this situation. CRBSI was defined as a combination of at least one positive peripheral blood culture sampled immediately before or within 48 h after catheter removal, colonization of the catheter tip with the same microorganism or a differential time to positivity of blood cultures ≥ 2 h, in the absence of any other infectious focus explaining the positive blood culture result.

Statistical analysis

No computation of sample size was performed a priori. Baseline characteristics of the patients and catheters included in the study were described using numbers (percentages) for categorical variables and median [inter-quartile range] for quantitative variables.

All the RCTs included in this work prospectively recorded demographics, clinical and microbiological data with a particular attention to risk factors associated with CVCs infection (Additional File 1). The statistical analysis plan had four steps that are detailed in the Additional File 2. The number of observations with missing data was low, and the imputation strategies are described in the Additional File 2.

Briefly, we identified independent risk factors of catheter-tip colonization from appropriate univariable and multivariable regression models in the training cohort [26]. We checked the absence of interaction between each risk factor and “CRI suspected” as the reason for CVC removal. Then, we assessed internal validity by resampling methods to retain only robust risk factors, and to estimate their coefficients to correct for over optimism [27]. Based on the strength of their association with the risk of catheter-tip colonization, we derived a 0–12 points score (called “CVC-OUT”) with a higher value signifying higher catheter-tip colonization probability [28]. Finally, the external validity of the CVC-OUT score was tested in the testing cohort. In both cohorts, the discrimination of the score was assessed using Receiving Operator Characteristics (ROC) curves and their c-index. The fitness was tested by the Hosmer and Lemeshow test. All analyses were performed with SAS software V9.4 (SAS Institute, NC, Cary) and R software (R Foundation for Statistical Computing, Vienna, Austria).


Patients and catheters

In the training cohort, 3471 CVCs in the 3SITES study and 897 DCs in the CATHEDIA study were available. Among these 4368 catheters, 469 were excluded due to missing data regarding the colonization status, representing a total of 3899 catheters included in the study, of which 3095 (79%) were CVCs and 804 (21%) dialysis catheters (Fig. 1). Baseline characteristics of patients and catheters are presented in Table 1. Most patients were male (65%), with a median age of 65 years. 845 patients (22%) had a body mass index (BMI) > 30 kg/m2, and 833 (21%) had diabetes. Data concerning BMI were missing for 225 (6%) patients, and data concerning a past medical history of diabetes were missing for 331 (9%) patients. 2289 patients (59%) received vasopressors and 3127 (80%) required mechanical ventilation. Data concerning vasopressors infusion at admission were missing for 336 (9%) patients. At insertion, 1,405 patients (36%) received anticoagulants and 2300 (59%) received antibiotic therapy. Forty percent (n = 1562), 39% (n = 1511) and 22% (n = 826) of catheters were inserted in the femoral, jugular and subclavian vein, respectively. Antiseptic used for skin disinfection was mostly iodine-povidone (58%). A total of 546 catheters (14%) were second or subsequent catheters. Median dwell time was 5 days.

Fig. 1
figure 1

Flowchart. 1CVC, Central Venous Catheter; 2DC, Dialysis Catheter

Table 1 Patients and catheters characteristics in the training cohort and in the testing cohort

In the testing cohort (Fig. 1), 6848 catheters were enrolled, of which 4119 (60%) CVCs and 2729 (40%) DCs. Patients were mostly male (64%), with a median age of 64 years. At admission, 5023 (73%) required mechanical ventilation, and 3756 (55%) received vasopressors. 4413 (64%) patients received antibiotics. 45% (n = 3087), 34% (n = 2320) and 21% (n = 1441) of catheters were inserted in the femoral, jugular and subclavian vein respectively.

The distribution of important covariates in the training and testing cohorts is available in Table 1.

In the training cohort, 575 (15%, 95% confidence interval (CI) [14,15,16]) catheters were colonized. Among 597 catheters removed for suspected CRI, 477 (80%, 95% CI [77–83]) were culture-negative. In the testing cohort, 588 (9%, 95% CI [8, 9]) catheters were colonized. Among 872 catheters removed for suspected CRI, 723 (83%, 95% CI [80–85]) were culture-negative.

The overall cohort represented 10,747 catheters, of which 1163 (11%, 95% CI [10, 11]) were colonized. Among 1469 catheters removed for suspected CRI, 1200 (82%, 95% CI [80–84]) were culture-negative.

Predictors of outcome

On univariable analysis, nine factors were significantly associated with colonization: age > 60 years, obesity, diabetes, mechanical ventilation, use of anticoagulants, site of insertion (femoral and jugular versus subclavian), type of catheter (DC versus CVC), successful insertion at first attempt and dwell time. The first catheter inserted was significantly less likely to be colonized (Table 2).

Table 2 Univariable and multivariable analyses

None of the risk factor interacted significantly with “suspected CRI” as the reason for CVC removal. No evidence of collinearity was observed (Additional file 3: Table S1), and all ten covariates were introduced in the multivariable model. After backward and forward selection, age > 60 years, obesity, diabetes, mechanical ventilation, site insertion, type of catheter (DC versus CVC), rank of the catheter (first versus subsequent) and dwell time > 5 days were significantly associated with colonization. In contrast, use of anticoagulants and successful insertion at first attempt were no longer associated with colonization on multivariable analysis (Table 2). The sensitivity analysis consisting in repeating the multivariable analysis in a five-time imputed dataset was similar (Additional file 4: Table S2).

Based on 500-time repetitions of the multivariable model in bootstrap samples and sub-samples, obesity, diabetes, site of insertion, type of catheter, rank of the catheter and dwell time were considered as robust risk factors for colonization (Additional file 5: Table S3). To derive the CVC-OUT score as a simple points-based score, points were attached to the predictors according to their bootstrap coefficients. The method used to calculate the points system is detailed in Additional file 6. Obesity, diabetes and rank of the catheter > 1 accounted for 1 point. Type of the catheter (DC) and dwell time > 5 days accounted for 2 points. Insertion site accounted for 4 or 5 points respectively in case of jugular or femoral insertion. The score ranged from 0 (predicted probability for colonization = 2.6%, 95% CI [1.9–3.4]) to 12 points (predicted probability for colonization = 50.6%, 95% CI [44.6–56.4]). The ROC curve is presented in Fig. 2. Area under curve (AUC) was 0.69, 95% CI [0.67–0.72], and the Hosmer and Lemeshow goodness of fit test denoted satisfying adequation to the data (p = 0.64). Using the ROC curve and the zenith of the Youden index to determine the threshold, a CVC-OUT score ≥ 6 points defined the “high risk” group and predicted colonization with a sensitivity of 79.5% (95% CI [75.9–82.7]) and a specificity of 49.0% (95% CI [47.2–50.7]). The positive predictive value (PPV) was 21.2% (95% CI [20.3–22.1]) and the negative predictive value (NPV) was 93.2% (95% CI [92.1–94.2]) (Additional file 7: Table S4). Comparison of the observed percentages of colonized catheters and the CVC-OUT score predicted probability in the training cohort is displayed in Additional file 8: Figure S1.

Fig. 2
figure 2

ROC curves for the simplified based-points score in the training, testing, and overall cohorts. *Threshold ≥ 6 points (“high risk group”) versus < 6 points (“low risk group”) a. shows the receiving operator comparison (ROC) curve in the training cohort, b. shows the ROC curve in the testing cohort and c. shows the comparison between a CVC-OUT score ≥ 6 points and the suspicion of catheter-related infection (CRI) in the overall cohort. AUC, area under curve; CI, confidence interval; CRI, catheter-related infection

External validation

In the testing cohort, the CVC-OUT score ranked from 0, predicting a colonization probability of 3.4% (95% CI [2.6–4.4]), to 12, predicting a colonization probability of 17.2% (95% CI [14.4–20.5]). AUC was 0.60, 95% CI [0.58–0.62], p < 0.001 (Fig. 2). The sensitivity was 79.1% (95% CI [75.6–82.3]) and the specificity was 33.2% (95% CI [32.1–34.4]). The use of CVC-OUT score < 6 points allowed excluding the diagnosis of colonization with a NPV of 94.4% (95% CI [93.5–95.2]) Additional file 7: Table S4. Comparison of the observed percentages of colonized catheters and the CVC-OUT score predicted probability in the testing cohort is displayed in Additional file 8: Figure S1.

Predictive performance in the overall cohort

The CVC-OUT score in the overall cohort ranked from 0 (predicted probability 3.3%, 95% CI [2.7–4.0]) to 12 points (predicted probability 26.2%, 95% CI [23.3–29.2]). Figure 3 displays the comparison between observed percentages of colonized CVCs and score-predicted probability for colonization by CVC-OUT score in the overall cohort.

Fig. 3
figure 3

Comparison of the observed percentage of colonized catheters and the score-predicted probability for colonization by points total in the overall cohort (n = 10,747)

AUC for the CVC-OUT score in the overall cohort was 0.63, 95% CI [0.61–0.64]. As shown in Table 3, the negative predictive value (NPV) of a CVC-OUT score < 6 points was 93.9%, 95% CI [93.2%–94.5%]. A score ≥ 6, predicted colonization better than the presence of a suspicion of CRI (p < 0.001 for the comparison of AUCs) (Fig. 2).

Table 3 Contingency table for a CVC-OUT score ≥ 6 in the overall cohort (n = 10,747)

Among catheters removed for suspected CRI (n = 1469), removing only those with a score ≥ 6 could avoid 402 unnecessary removals and reinsertions (28.6%, 95% CI [26.3–30.9]) at the cost of 59 colonized catheters left in situ (4.0%, 95% CI [3.0–5.0]). Among the catheters removed for suspected CRI in the overall cohort, AUC for the simplified points score was 0.59, 95% CI [0.55–0.62] (Additional file 9: Figure S2).

Sensitivity analysis

After exclusion of catheters colonized with coagulase-negative Staphylococci (CoNS), 357 out of 3,861 (10%, 95% CI [9,10,11]) catheters in the training cohort and 345 out of 6396 (5%, 95% CI [5, 6]) catheters in the testing cohort were colonized. Age, obesity, diabetes, insertion site (jugular and femoral vs subclavian), type of catheter (CVC vs. DC) and dwell time > 5 days were significantly associated with colonization and proved robust after the bootstrap selection (Additional file 10: Tables S4 and S5).


Using the high-quality individual-patient data from five large RCTs investigating the prevention of CRI in the ICU, we developed and validated a score with a moderate discrimination but a good negative predictive value to predict catheter-tip colonization. The presence of a low CVC-OUT score may encourage physicians to adopt a watchful waiting strategy that is safer while remaining CVCs in place in case of suspicion of CRI.

The points system makes the score easy to use at the bedside and allows considering the weight of each risk factor. In our study, the femoral and the jugular veins, respectively 5 and 4 points were the factors most strongly associated with colonization. Interestingly, anatomical insertion site is modifiable. A prospective study by Timsit et al. identified femoral and internal jugular sites of insertion and dwell time ≥ 5 days as independent risk factors for catheter-tip colonization [29]. In our study, dwell time was also associated with colonization with the odds increasing almost twofold when the catheter was left in place more than five days. This finding suggests that it is appropriate to favor subclavian access when the expected duration of catheterization is long, as supported by recent guidelines [30]. Increasing the rate of subclavian CVCs inserted in the ICU will not only reduce the risk of intravascular complications [3] but will also provide reassurance that any potential infection symptoms occurring during the ICU stay are less likely to be catheter-related. Of note, the subclavian route is associated with a higher risk of mechanical complications, including pneumothorax [3]. The choice of the insertion site must therefore remain based on an individual assessment of the benefit-risk ratio.

The proportion of femoral CVCs was high in our study, about 43%. Two reasons could explain this: (i) DCs represented 21% and 40% of the training and testing cohorts, respectively, limiting the choice of insertion site to the jugular and femoral veins. The CATHEDIA RCT did not support to avoid femoral central venous access for DCs, except amongst obese patients [20]. Second, the 3SITES RCT investigated the anatomic site for CVC insertion in a 1:1:1 ratio, so a third of participants had femoral CVCs [3].

Consistent with our observation, a recent study showed that BMI ≥ 40 kg/m2 was associated with catheter colonization [31]. This could be explained by more frequent dressing disruptions, and by insertion made more difficult by the loss of anatomical landmarks and deeper vessels, thus favoring more manipulations [32].

Most studies focus on CRI or CRBSI and few have investigated the risk factors for colonization [29, 31, 33]. The clinical relevance of colonization (regardless of the presence of CR-1 or CR-2) is debated as a surrogate endpoint for CRBSI [34, 35] but it represents the first step to CRI [36]. The clinical context of CRI suspicion (fever or local signs of inflammation in presence of a CVC) provides more importance to catheter-tip colonization, as it naturally excludes asymptomatic patients. Of note, CRI is a moderate to good surrogate endpoint for CRBSI [34, 35]. Predicting colonization risk in this context, or more importantly its low likelihood with high negative predictive value could support the “watchful waiting” strategy, and help reduce the prevalence of non-colonized catheters being removed.

As we focused on colonization, we performed a sensitivity analysis by excluding catheters for which the tip culture was positive to CoNS only. Although CoNS are frequently associated with CRBSI [1, 37], in the absence of positive blood cultures, identification of CoNS on catheter-tip culture is difficult to interpret. The continuum between contamination and colonization with pathogenic potential is blurred, as CoNS tends to be less virulent [38], and may be less likely to cause local or systemic symptoms [39].

Our study has several limitations. We investigated colonization, which is not a clinical outcome such as CRI. Therefore, our findings do not support CVC removal in asymptomatic patients who have a high CVC-OUT score without another indication for removing the CVC. Because of its retrospective design, unmeasured covariates and missing data may cause residual confounding even though data collection was prospective in all five RCTs, and performed by trained investigators. Data regarding antibiotic therapy at catheter removal were not available, and it may have affected the results of catheter-tip cultures. Indeed, a study by Souweine et al. raised awareness that antibiotics administrated at the time of catheter removal may interfere with microbiological diagnosis [40]. All five studies were conducted before the current French Guidelines, and practice has evolved since then. It has now been clearly demonstrated that skin disinfection with alcoholic chlorhexidine is more effective than povidone iodine to prevent CRI [6, 21, 22]. The difference in the prevalence of colonization in the training and testing cohorts may be partially explained by a higher prevalence of skin disinfection with 2% chlorhexidine alcohol in the testing cohort. We chose not to adjust the multivariable model on the type of skin antiseptic used, because using alcoholic 2% chlorhexidine is now recommended in most guidelines [10, 11, 30]. This may have artificially decreased the discrimination of the CVC-OUT score by omitting the skin antisepsis or dressings used in the trials. For the same reason, we did not adjust the multivariable model on the use of ultrasonography guidance at catheter insertion, even though a recent study showed that the use of ultrasonography might favor asepsis breaches and thus increase the risk of CRI [41]. Of note, the use of evidence-based prevention strategies that reduce the risk of CRI will tend to increase our negative predictive value by decreasing the incidence of catheter-tip colonization. Our study included both: CVCs and DCs. They differ in design, use (with more manipulations for DCs) and insertion site [42]. The subclavian vein, while preferred to reduce the CVC infection rate, is discouraged for DCs insertion because of the risk of central-vein stenosis [43]. Finally, the potential contribution of the immunocompromised state of those patients needing dialysis or more frequent manipulations of the central line may explain why DC were at higher risk of colonization, compared to CVCs. Because the interventions differed within the studies, there could be an interaction between the intervention and the outcome. Nevertheless, studies included in the training cohort (3SITES and CATHEDIA) both investigated the impact of insertion site on catheter-related complications, with randomization of the insertion site, thus limiting heterogeneity within the training cohort. Finally, the five multicenter RCTs were conducted exclusively in France, which potentially limits the generalizability of our results.

In conclusion, the CVC-OUT score showed a moderate ability to discriminate catheter-tip colonization but a high negative predictive value and allowed to apprehend the weight of each risk factor. Preference of the subclavian site was the strongest and only modifiable factor influencing the risk of CVC-tip colonization. Of note, the use of real-time ultrasound decreases the risk of mechanical complications in subclavian insertions as compared to landmark technique, without being associated with an increased risk of infection [41]. Whether the use of CVC-OUT is effective to reduce unnecessary CVC removal in the clinical context of suspected CRI requires further prospective evaluations.

Availability of data and materials

The datasets used and analyzed during the current study are available from the corresponding author on reasonable request.



Central venous catheter


Catheters used for renal replacement therapy


Intensive care unit


Catheter-related bloodstream infection


Catheter-related infection


Randomized controlled trial


European Centre for Disease Prevention and Control

ROC curve:

Receiver operating characteristics curve


Confidence interval


Odds ratio


Area under curve


Positive predictive value


Negative predictive value


Coagulase-negative Staphylococci


  1. European Centre for Disease Prevention and Control. Healthcare-associated infections in Intensive Care Units. Annual epidemiological report; 2017.

  2. Timsit JF, Rupp M, Bouza E, Chopra V, Kärpänen T, Laupland K, et al. A state of the art review on optimal practices to prevent, recognize, and manage complications associated with intravascular devices in the critically ill. Intensive Care Med. 2018;44(6):742–59. (PMID: 29754308).

    Article  PubMed  Google Scholar 

  3. Parienti JJ, Mégarbane B, Kalfon P, Thuong M, Savary B, Terzi N, et al. Intravascular complications of central venous catheterization by insertion site. N Engl J Med. 2015. (PMID: 26398070).

    Article  PubMed  Google Scholar 

  4. Günther SC, Schwebel C, Hamidfar-Roy R, Bonadona A, Lugosi M, Ara-Somohano C, et al. Complications of intravascular catheters in ICU: definitions, incidence and severity. A randomized controlled trial comparing usual transparent dressings versus new-generation dressings (the ADVANCED study). Intensive Care Med. 2016;42(11):1753–65. (PMID: 27734108).

    Article  PubMed  Google Scholar 

  5. Ista E, van der Hoven B, Kornelisse RF, van der Starre C, Vos MC, Boersma E, et al. Effectiveness of insertion and maintenance bundles to prevent central-line-associated bloodstream infections in critically ill patients of all ages: a systematic review and meta-analysis. Lancet Infect Dis. 2016;16(6):724–34. (PMID: 26907734).

    Article  PubMed  Google Scholar 

  6. Pages J, Hazera P, Mégarbane B, du Cheyron D, Thuong M, 3SITES Study Group, et al. Comparison of alcoholic chlorhexidine and povidone–iodine cutaneous antiseptics for the prevention of central venous catheter-related infection: a cohort and quasi-experimental multicenter study. Intensive Care Med. 2016;42(9):1418–26. (PMID: 27311311).

    CAS  Article  PubMed  Google Scholar 

  7. Prowle JR, Echeverri JE, Ligabo EV, Sherry N, Taori GC, Crozier TM, et al. Acquired bloodstream infection in the intensive care unit: incidence and attributable mortality. Crit Care. 2011;15(2):R100. (PMID: 21418635).

    Article  PubMed  PubMed Central  Google Scholar 

  8. Blot SI, Depuydt P, Annemans L, Benoit D, Hoste E, De Waele JJ, et al. Clinical and economic outcomes in critically ill patients with nosocomial catheter-related bloodstream infections. Clin Infect Dis. 2005;41(11):1591–8. (PMID: 16267731).

    Article  PubMed  Google Scholar 

  9. Ziegler MJ, Pellegrini DC, Safdar N. Attributable mortality of central line associated bloodstream infection: systematic review and meta-analysis. Infection. 2015;43(1):29–36. (PMID: 25331552).

    Article  PubMed  Google Scholar 

  10. Mermel LA, Allon M, Bouza E, Craven DE, Flynn P, O’Grady NP, et al. Clinical practice guidelines for the diagnosis and management of intravascular catheter-related infection: 2009 update by the infectious diseases society of America. Clin Infect Dis. 2009;49(1):1–45. (PMID: 19489710).

    CAS  Article  PubMed  Google Scholar 

  11. Timsit JF, Baleine J, Bernard L, Calvino-Gunther S, Darmon M, Dellamonica J, et al. Expert consensus-based clinical practice guidelines management of intravascular catheters in the intensive care unit. Ann Intensive Care. 2020;10(1):118. (PMID: 32894389).

    Article  PubMed  PubMed Central  Google Scholar 

  12. Rijnders BJ, Peetermans WE, Verwaest C, Wilmer A, Wijngaerden EV. Watchful waiting versus immediate catheter removal in ICU patients with suspected catheter-related infection: a randomized trial. Intensive Care Med. 2004;30(6):1073–80. (PMID: 14999442).

    Article  PubMed  Google Scholar 

  13. Lorente L, Martín MM, Vidal P, Rebollo S, Ostabal MI, Vascular Catheter Working Group of GTEIS/SEMICYUC, et al. Should central venous catheter be systematically removed in patients with suspected catheter related infection? Crit Care. 2014;18(5):564. (PMID: 25514404).

    Article  PubMed  PubMed Central  Google Scholar 

  14. Burnham JP, Rojek RP, Kollef MH. Catheter removal and outcomes of multidrug-resistant central-line-associated bloodstream infection. Medicine. 2018;97(42): e12782. (PMID: 30334966).

    Article  PubMed  PubMed Central  Google Scholar 

  15. Jensen AG, Wachmann CH, Espersen F, Scheibel J, Skinhøj P, Frimodt-Møller N. Treatment and outcome of Staphylococcus aureus bacteremia: a prospective study of 278 cases. Arch Intern Med. 2002;162(1):25. (PMID: 11784216).

    CAS  Article  PubMed  Google Scholar 

  16. Calò F, Retamar P, Martínez Pérez-Crespo PM, Lanz-García J, Sousa A, Goikoetxea J, et al. Catheter-related bloodstream infections: predictive factors for Gram-negative bacteria aetiology and 30 day mortality in a multicentre prospective cohort. J Antimicrob Chemother. 2020;75(10):3056–61. (PMID: 32688386).

    CAS  Article  PubMed  Google Scholar 

  17. Zhong Y, Deng L, Zhou L, Liao S, Yue L, Wen SW, et al. association of immediate reinsertion of new catheters with subsequent mortality among patients with suspected catheter infection: a cohort study Ann Intensive Care. 2022;12(1):38. (PMID: 35524924).

    Article  PubMed  Google Scholar 

  18. Buetti N, Ruckly S, Lucet JC, Bouadma L, Garrouste-Orgeas M, Schwebel C, et al. Local signs at insertion site and catheter-related bloodstream infections: an observational post hoc analysis using individual data of four RCTs. Crit Care. 2020;24(1):694. (PMID: 33317594).

    Article  PubMed  PubMed Central  Google Scholar 

  19. Collins GS, Reitsma JB, Altman DG, Moons KGM. Transparent Reporting of a Multivariable Prediction Model for Individual Prognosis Or Diagnosis (TRIPOD): the TRIPOD statement. J Clin Epidemiol. 2015;68(2):112–21. (PMID: 25569120).

    Article  Google Scholar 

  20. Parienti JJ, Thirion M, Mégarbane B, Souweine B, Ouchikhe A, Polito A, et al. Femoral vs jugular venous catheterization and risk of nosocomial events in adults requiring acute renal replacement therapy: a randomized controlled trial. JAMA. 2008;299(20):2413. (PMID: 18505951).

    CAS  Article  PubMed  Google Scholar 

  21. Mimoz O, Lucet JC, Kerforne T, Pascal J, Souweine B, Goudet V, et al. Skin antisepsis with chlorhexidine–alcohol versus povidone iodine–alcohol, with and without skin scrubbing, for prevention of intravascular-catheter-related infection (CLEAN): an open-label, multicentre, randomised, controlled, two-by-two factorial trial. Lancet. 2015;386(10008):2069–77. (PMID: 26388532).

    CAS  Article  PubMed  Google Scholar 

  22. Timsit JF, Mimoz O, Mourvillier B, Souweine B, Garrouste-Orgeas M, Alfandari S, et al. Randomized controlled trial of chlorhexidine dressing and highly adhesive dressing for preventing catheter-related infections in critically ill adults. Am J Respir Crit Care Med. 2012;186:7. (PMID: 23043083).

    CAS  Article  Google Scholar 

  23. Souweine DB, Lautrette DA, Gruson D, Canet DE, Klouche K, Argaud L, et al. Ethanol lock and risk of hemodialysis catheter infection in critically ill patients: a randomized controlled trial. Am J Respir Crit Care Med. 2015. (PMID: 25668557).

    Article  PubMed  Google Scholar 

  24. Brun-Buisson C, Abrouk F, Legrand P, Huet Y, Larabi S, Rapin M. Diagnosis of central venous catheter-related sepsis critical Level of quantitative tip cultures. Arch Intern Med. 1986;147(5):873–7 (PMID: 3555377).

    Article  Google Scholar 

  25. European Centre for Disease Prevention and Control. European surveil ance of healthcare- associated infections and prevention indicators in intensive care units—HAI-Net ICU protocol, version 2.2. Stockholm: ECDC. 2017.

  26. Liang KY, Zeger SL. Longitudinal data analysis using generalized linear models. Biometrika. 1986;73(1):13–22.

    Article  Google Scholar 

  27. Mannan H. A practical application of a simple bootstrapping method for assessing predictors selected for epidemiologic risk models using automated variable selection. Int J Stat Appl. 2017.

    Article  Google Scholar 

  28. Sullivan LM, Massaro JM, D’Agostino RB. Presentation of multivariable data for clinical use: the Framingham study risk score functions. Statist Med. 2004;23(10):1631–60. (PMID: 15122742).

    Article  Google Scholar 

  29. Timsit JF, L’Hériteau F, Lepape A, Francais A, Ruckly S, Venier AG, et al. A multicentre analysis of catheter-related infection based on a hierarchical model. Intensive Care Med. 2012;38(10):1662–72. (PMID: 22797354).

    CAS  Article  PubMed  Google Scholar 

  30. Buetti N, Marschall J, Drees M, Fakih MG, Hadaway L, Maragakis LL, et al. Strategies to prevent central line-associated bloodstream infections in acute-care hospitals: 2022 Update. Infect Control Hosp Epidemiol. 2020. (PMID: 35437133).

    Article  Google Scholar 

  31. Buetti N, Souweine B, Mermel L, Mimoz O, Ruckly S, Loiodice A, et al. Obesity and risk of catheter-related infections in the ICU. A post hoc analysis of four large randomized controlled trials. Intensive Care Med. 2021;47(4):435–43. (PMID: 33521871).

    Article  PubMed  Google Scholar 

  32. Dossett LA, Dageforde LA, Swenson BR, Metzger R, Bonatti H, Sawyer RG, et al. Obesity and site-specific nosocomial infection risk in the intensive care unit. Surg Infect. 2009;10(2):137–42. (PMID: 19388836).

    Article  Google Scholar 

  33. Moro ML, Viganò EF, Lepri AC. The central venous catheter-related infections study group. Risk factors for central venous catheter-related infections in surgical and intensive care units. Infection Control Hospital Epidemiol. 1994;15(4):253–64 (PMID: 8207192).

    CAS  Article  Google Scholar 

  34. Rijnders BJA, Van Wijngaerden E, Peetermans WE. Catheter-tip colonization as a surrogate end point in clinical studies on catheter-related bloodstream infection: how strong is the evidence? Clin Infect Dis. 2002;35(9):1053–8. (PMID: 12384838).

    Article  PubMed  Google Scholar 

  35. de Grooth HJ, Timsit JF, Mermel L, Mimoz O, Buetti N, du Cheyron D, et al. Validity of surrogate endpoints assessing central venous catheter-related infection: evidence from individual- and study-level analyses. Clin Microbiol Infect. 2020;26(5):563–71. (PMID: 31586658).

    Article  PubMed  Google Scholar 

  36. Safdar N, Maki DG. The pathogenesis of catheter-related bloodstream infection with noncuffed short-term central venous catheters. Intensive Care Med. 2004;30(1):62–7. (PMID: 14647886).

    Article  PubMed  Google Scholar 

  37. Buetti N, Lo Priore E, Atkinson A, Widmer AF, Kronenberg A, Marschall J. Catheter-related infections: does the spectrum of microbial causes change over time. A nationwide surveillance study. BMJ Open. 2018;8(12):e023824. (PMID: 30580270).

    Article  PubMed  PubMed Central  Google Scholar 

  38. von Eiff C, Peters G, Heilmann C. Pathogenesis of infections due to coagulasenegative staphylococci. Lancet Infect Dis. 2002;2(11):677–85. (PMID: 12409048).

    Article  Google Scholar 

  39. Safdar N, Maki DG. Inflammation at the insertion site is not predictive of catheter-related bloodstream infection with short-term, noncuffed central venous catheters. Crit Care Med. 2002;30(12):2632–5. (PMID: 12483050).

    Article  PubMed  Google Scholar 

  40. Souweine B, Heng AE, Aumeran C, Thiollière F, Gazuy N, Deteix P, et al. Do antibiotics administered at the time of central venous catheter removal interfere with the evaluation of colonization? Intensive Care Med. 2008;34(2):286–91. (PMID: 17882399).

    Article  PubMed  Google Scholar 

  41. Buetti N, Mimoz O, Do LM, Ruckly S, Dupuis C, Mira JP, et al. Ultrasound guidance and risk for central venous catheter-related infections in the ICU.A post hoc analysis of individual data of three multi-centric randomized trials. Clin Infect Dis. 2021;73(5):1054–61. (PMID: 33277646).

    Article  Google Scholar 

  42. Schneider A, Baldwin I, Souweine B. What’s new: prevention of acute dialysis catheter-related infection. Intensive Care Med. 2018;44(3):356–8. (PMID: 28523404).

    Article  PubMed  Google Scholar 

  43. Boussekey N, Chiche A, Faure K. Section 5: dialysis Interventions for treatment of AKI. Kidney Int Suppl. 2012;2(1):89–115.

    Article  Google Scholar 

Download references


We thank all contributing investigators of the 3SITES, CATHEDIA, CLEAN, DRESSING-2 and ELVIS trials.


JI received a grant (“Bourse Année Recherche”) from the Caen University Hospital; NB received a grant from the Swiss National Science Foundation (grant number: P4P4PM_194449). CATHEDIA was funded by the Centre Hospitalier Universitaire de Caen, Caen, France, and supported by an unrestricted academic grant from the French Health Ministry (Programme Hospitalier de Recherche Clinique National 2003). 3SITES was supported by a grant from the Hospital Program for Clinical Research, French Ministry of Health (PHRC-N 2010, 06-03). CLEAN was supported by the University Hospital of Poitiers and CareFusion. DRESSING-2 was supported by University of Grenoble and 3 M Company. ELVIS was supported by grants from the French Ministry of Health (Programme Hospitalier de Recherche Clinique National 2008 “ELVIS” and Programme Hospitalier de Recherche Clinique Inter-regional 2009 “ESCAPE”).

Author information

Authors and Affiliations



JJP, NB and JFT designed the study. OM, BS, JJP and JFT acquired the data. JI, AB, JJP, NB and HJdG did the statistical analysis. JI, JJP, NB, JFT, HdJG, LM and SR analyzed and interpreted the data. JI, JJP and AB drafted the manuscript. NB, HJdG, OM, BM, JPM, BS, DdC, LM and JFT substantively revised the manuscript. All authors reviewed the manuscript and approved the final report.

Corresponding author

Correspondence to Jean-Jacques Parienti.

Ethics declarations

Ethics approval and consent to participate

All procedures performed in studies were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki Declaration and its later amendments or comparable ethical standards. The studies were approved by the national ethic committees. Further ethical consent was not required according to the French law for research.

Consent for publication

Not applicable.

Competing interests

JI no conflict; NB no conflict; HJdG. no conflict; AB no conflict; OM received research grants, lecture and consultancy fees from CareFusion; BM no conflict; JPM has no conflict directly related to the article but served on the scientific board of MSD, LFB and ASAI and Astellas; SR no conflict; B.S no conflict; DdC. no conflict; LM served as an advisor/consultant for Citius Pharmaceuticals, Marvao Medical, Leonard-Meron Biosciences, Destiny Pharma and Nobio; JFT. has no conflict directly related to the article but serves on the advisory boards of MSD, Pfizer, Bayer pharma and Nabriva, has given lectures at MSD, Pfizer, Biomerieux, and his university or research team had received research grants from Pfizer and MSD; JJP. no conflict.

Additional information

Publisher's Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary Information

Additional file 1

Annex 1: Detailed presentation of the randomized controlled trials included in the study: Text describing inclusion/exclusion criteria and interventions for the five trials included in the study.

Additional file 2

Annex 2: Detailed Statistical Methods: Text describing the statistical analysis.

Additional file 3

Table S1: Variance Inflation Factors for the covariates included in the multivariable model in the principal analysis.

Additional file 4

Table S2: Univariable and multivariable analyses in the five time imputed dataset as part of a sensitivity analysis.

Additional file 5

Table S3: Relative frequency (%) with which each candidate predictor was selected in 500 bootstrap samples from the training cohort and correspondence between the bootstrapped coefficients for the robust risk factors and attached points in the simplified points-based score.

Additional file 6

Annex 3: Method for the determination of the attached points in the points-based system. Description of the process of derivating points from the coefficients.

Additional file 7

Table S4: Contingency tables for the CVC-OUT score in the training cohort and in the testing cohort.

Additional file 8

Figure S1: Comparison of the observed percentage of colonized catheters and the score-predicted probability for colonization by points total in the training and testing cohorts panel a. displays the training cohort and panel b. displays the testing cohort.

Additional file 9

Figure S2: ROC curve for the simplified points-based score in the sub-group of catheters removed for suspected catheter-related infection.

Additional file 10

Table S4: Univariable and multivariable analyses after exclusion of the catheters colonized with Coagulase negative Staphylococci as part of a sensitivity analysis. Table S5: Relative frequency (%) with which each candidate predictor was selected in 500 bootstrap samples from the training cohort and bootstrapped coefficients for the robust risk factors after exclusion of the catheters colonized with Coagulase negative Staphylococci as part of a sensitivity analysis.

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit The Creative Commons Public Domain Dedication waiver ( applies to the data made available in this article, unless otherwise stated in a credit line to the data.

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Iachkine, J., Buetti, N., de Grooth, HJ. et al. Development and validation of a multivariable prediction model of central venous catheter-tip colonization in a cohort of five randomized trials. Crit Care 26, 205 (2022).

Download citation

  • Received:

  • Accepted:

  • Published:

  • DOI:


  • Catheter-tip colonization
  • Catheter-related infection
  • Intensive care unit
  • Central Venous catheters
  • Predictive score