Risk Factors For Acute Adverse Events During Ultrasound Guided Central Venous Cannulation in the Emergency Department

Abstract

Ultrasound (US) greatly facilitates cannulation of the internal jugular vein. Despite the ability to visualize the needle and anatomy, adverse events still occur. We hypothesized that the technique has limitations among certain patients and clinical scenarios. Objective: The purpose of our study was to identify characteristics of adverse events surrounding US-guided central venous cannulation (CVC). Methods: We assembled a prospective observational cohort of emergency department (ED) patients undergoing consecutive CVC under US. The primary outcome of interest was a composite of acute mechanical adverse events including hematoma, arterial cannulation, pneumothorax, and unsuccessful placement. Physicians performing the CVC recorded anatomical site, reason for insertion, and acute complications. We followed the catheters until they were removed based on radiographic evidence hospital nursing records. ED charts and pharmacy records contributed variables of interest. A self reported online survey provided physician experience information. Logistic regression was used to calculate the odds of an adverse outcome. Results: Physicians attempted 289 CVC on 282 patients. An adverse outcome occurred in 57 attempts (19.7%, 95%CI 15.5–24.7) the most common being 31 unsuccessful placements (11%, 95% CI 7.7–14.8). Patients with a history of end stage renal disease (OR 3.54, 95%CI 1.59–7.89) and central lines placed by operators with intermediate experience (OR 2.26, 95% CI 1.19–4.32) were most likely to encounter adverse events. Previously cited predictors such as BMI, coagulopathy, and pulmonary hyperinflation were not significant in our final model. Conclusions: Acute adverse events occurred in approximately 20% of US guided central line attempts. We identified both patient (history of end stage renal disease) and physician (intermediate experience level) factors that are associated with acute adverse events.

Introduction

Prior to the use of ultrasound (US) technology, several observational studies conducted in settings outside of the Emergency Department (ED) reported that acute adverse events occur in 6.3–11.8% of all central line placement attempts. ^1,2 A single study based solely on ED attempts reported rates as low as 3.4%.³ US guidance reportedly decreases adverse events associated with the procedure. In several settings, US assistance decreased by half the mean number of attempts required to cannulate the vessel and resulted in significant time reduction of the procedure. ^4–8 Two randomized controlled trials in the ED also corroborate the benefit of US guidance reporting success rates as high as 98% as well as a 10% absolute decrease of acute mechanical adverse events. ^{9, 10} In addition to guiding operators, US can reveal anatomic variations, thrombi, and damaged vessels. ^{11, 12} However, despite these successes, the procedure still fails in some patients.^{9, 13,10} Body Mass Index (BMI), operator experience, coagulopathy, and history of prior cannulation have been implicated as factors that might increase the chances of an acute adverse event during vascular access without US guidance.^14–16 The goal of this study is to better identify patient and operator characteristics that might predispose patients to acute mechanical adverse events from central venous cannulations (CVC) when using US.

Methods

Study Design

This was a prospective observational study of consecutive central line attempts using US guidance to cannulate the internal jugular vein in the emergency department (ED).

Setting

The study was conducted in a level 1, academic, tertiary-care referral center with a volume of 84,000 visits per year between March 2007 and March 2008. The 4-year program in emergency medicine consists of 48 residents (12 per class) and 51 faculty. At the time of this study, emergency medicine US directors taught all residents and faculty 2 one hour courses on US-guided central line techniques in lecture format. The lectures consisted of video demonstrating internal jugular vein location, positioning the transducer and US unit, and common pitfalls for cannulation including anatomical variations and thrombosis. No simulation sessions were offered. Physicians further developed their skills in the “apprenticeship model” meaning they gained experience while managing and caring for patients who required central venous cannulation (CVC) in the presence of those familiar with the procedure. During this time period several attending physicians used US guidance but there was no formal evaluation process of their skill.

Selection of Participants

Enrolled subjects included a consecutive sample of patients who required US-guided CVC as determined by the team of ED physicians treating the patient. Patients with concurrent central venous access of the ipsilateral side, evidence of trauma to the neck or thoracic vessels, or any overlying dermal condition that posed a possible contraindication to the procedure were excluded. The Human Research Protection Office approved the study. Waiver of consent was granted.

Methods of Measurement

In conjunction with the Division of Emergency Medicine’s information technology (IT) section, we created a standardized electronic procedure note template. The procedure note was created for the purposes of recording outcomes of central venous cannulation attempts. Required form fields included the indication for insertion, the site, whether or not the operator used US guidance and any acute mechanical event that may have occurred during the procedure. Patients requiring venous cannulation due to lack of peripheral access were defined as “poor access.” Patients who required cannulation for active resuscitation due to shock or hemodynamic monitoring due to sepsis were defined as “septic/hypovolemic or hypotensive.” An attempt at cannulation during cardiopulmonary arrest secondary to trauma or cardiac arrest was defined as a “code line.” For the purposes of this study, we did not record how many needle passes were required for successful cannulation since our data suggested significant variability in reporting.

We defined US guidance as an attempt during which real-time US assisted the operator in guiding the needle towards the targeted vessel. Attempts that only used US to “mark” or “identify” the vessel but not actively guide the needle (the static technique) were not included in the study.

We defined an “attempt” as a single operator’s performance of the entire procedure from beginning to end (adequate aspiration and flush of all catheter ports) at a single anatomic site. Any needle passes by another operator, whether at the original site or at another site, was considered a separate attempt. We collected demographic and physical characteristics of the patients from the ED chart or hospital record. If no information regarding the patient’s height and weight were available, we collected this information from the patient’s driver’s license or other identification card. The Internationalized Normalized Ratio (INR) was obtained from the ED record. Hypotension in the ED was defined as any patient with a SBP ≥ 90 mm Hg. In addition, we captured the administration of vasopressors or inotropes as a surrogate measure for persistent hypotension despite resuscitation. Pharmacy records were cross-referenced with the patient’s visit date and patient name. Any patient who was administered epinephrine, norepinephrine, dopamine, or dobutamine was identified as requiring pressors and “persistently hypotensive.” After primary collection of the data related to the procedure, we retrospectively reviewed hospital charts on those who underwent ED cannulation for any history of renal disease requiring hemodialysis.

Emergency medicine faculty and residents inserted all central lines. Physician experience was defined by post-graduate year and self-reported cannulation experience. PGY-4 residents were grouped with members of the EM faculty since they could function as attending physicians in other settings. Operators completed a survey during the study in which they documented the number of central lines they had placed during their career and how many of their last 10 non-emergent central lines were inserted by US.

Outcome measures

The primary outcome of interest was a composite of acute mechanical adverse events attributed to central venous access. From review of the literature and consensus among our group we defined an adverse event as a cannulation attempt that resulted in a common complication or failure in completing the intended procedure.^16–21 Arterial puncture, pneumothorax, hematoma, extravascular or misplacement, and unsuccessful cannulation were all defined as acute adverse events. An arterial puncture occurred if the operator aspirated pulsatile arterial (bright red) blood into an 18-gauge locator syringe. We diagnosed pneumothorax/hemothorax by using plain radiographic films of the chest (either standard PA or portable) as read by a radiologist blinded to the technique used for central line cannulation. A pneumothorax/hemothorax identified within 24 hours of line placement was considered an adverse event unless the patient underwent another procedure prior to the CVC that could also cause a pneumothorax on the side in question. Operators or nurses involved in the procedure identified a hematoma if an area greater than 3 cm of bruising or swelling was noted around the insertion. Extravascular placement or misplacement was defined by failure to aspirate blood through any of the ports or a visible extravasation of intravenous fluid in the soft tissues as determined by the physician or nurse involved in the patient’s care. Unsuccessful cannulation was specifically defined as failure to complete the intended procedure by the primary operator for lack of successfully cannulating the intended vessel and aspirating through a port. Any alleged puncture of the vein that resulted in the inability to pass a guide wire was considered unsuccessful.

In cases where more than one adverse event occurred, we recorded the complication that required a therapeutic intervention as the primary adverse outcome. Pneumothorax/hemothorax, unsuccessful cannulation, and misplacement were considered primary adverse outcomes if they occurred simultaneously with arterial puncture and hematoma. Any occurrence of any complication was considered an “adverse event.”

In order to obtain data on delayed adverse events related to an ED central line, we followed the central lines placed in ED patients until evidence from either hospital records or radiologic records indicated that the ED central line was removed or replaced. This data was obtained retrospectively by three methods: review of cases involving CVC adverse events reported to the ED physician responsible for all quality improvement issues in the division, retrospective chart review of all nursing charts, and retrospective review of patient discharge summaries.

Discharge summaries were reviewed for any mention of “central line complication” or “cannulation device complication” or “iatrogenic” complication. Mention of any triggered an audit of inpatient nursing and operative records for the mention of adverse events related to an ED central line. Operative records were reviewed for any mention of a vascular procedure related to central line complications.

Primary data analysis

Results were entered into a Microsoft Excel (Redmond, WA) file and later transferred into a SAS (Version 9.2, Cary, NC) database. Normally distributed continuous variables were presented with means and standard deviations and non-normally distributed variables were presented as medians with interquartile range (IQR). To test associations with having an adverse event, the Chi-square test or Fisher’s exact test were used as appropriate for categorical variables. The t-test was used for normally distributed variables and the Wilcoxon-Mann- Whitney test was used for non-normally distributed variables.

All variables in this analysis were decided upon a priori based on findings in previous studies and by consensus among the group. Any variable on bivariate testing with a p value <.20 was eligible for the final model. We used this cutoff to allow sufficient opportunity to include patient and physician factors which might influence other variables included in the final model.

Since no explicit classification of “experience” exists, we defined operator experience based on similarly published studies that relied on the number of central line cannulations that physicians placed throughout their career. ^{9, 10} We created three levels of operator skill by grouping operators into three experience categories: those who, over the course of their career, had placed fewer than 20 central lines, those who had placed 20–50 central lines, and those who had placed greater than 50 central lines. We also categorized the likelihood of operators to place catheters using US guidance. Those who placed 8 of their last 10 central lines using US were compared to those who used US less often.

To identify variables independently associated with an adverse event, we performed multivariable logistic regression. The model was built in forward stepwise fashion and variables with a P>.05 were removed. In our final logistic model, missing physician data for US experience level (5%) was imputed and analyzed. Our results were similar so we present the final model without the imputed variables. We tested for interactions between variables in the final model. We used the Hosmer-Lemeshow goodness-of-fit test as diagnostic measures of our final model.

Results

The cohort consisted of 289 central venous cannulation attempts on 282 patients who underwent internal jugular cannulation under US guidance. Physicians indicated they placed 203 (70%, 95% CI 64.7 to 75.2) central lines for hypovolemia or sepsis resuscitation, 74 (26%, 95% CI 20.9 to 30.9) for lack of obtaining peripheral access and 12 (4%, 95% CI 2.4 to 7.1) during active cardiopulmonary resuscitation (ACLS/ATLS). The median age of patients undergoing cannulation was 60 years (IQR 49 to 75 years). The median BMI was 27.02 (IQR 22.63 to 32.27). A total of 187 patients exhibited one episode of hypotension (65%, 95% CI 59.0 to 70.0) and 102 (36%, 95% CI 30.0 to 50.0) patients required vasopressors for persistent hypotension. During our retrospective chart review for patients with a history of end stage renal disease and history of dialysis double data entry was performed. The kappa for determining a history of dialysis was .82 (95% CI .70 to .93). Our review identified 33 (11%, 95% CI 8.3 to 16.0) attempts on patients with a history of end stage renal disease.

A total of 77 physicians participated in the study. Attendings and residents in their last year of training performed 93 attempts (32%, 95% CI 27.1 to 37.8), PGY-3s performed 71 attempts (26%, 95% CI 20.0 to 29.9), PGY-2s performed 49 attempts (17%, 95% CI 13.1 to 21.7) and PGY-1s performed 76 attempts (25%, 95% CI 21.6 to 31.7). Physicians with fewer than 20 CVC insertions in their career attempted 34 (12%, 95% CI 8.5 to 16.0) cannulations and those with 20–50 career cannulations contributed 80 (29%, 95% CI 22.8 to 33.1) attempts. Physicians with over 50 career cannulations contributed 163 (59%, 95% CI 50.6 to 62.0) attempts. Information on operator experience was missing in 12 of the 289 attempts (4%). Among these, 7 (58%, 95% CI 32.0 to 80.7) attempts were performed by physicians in their first PGY year and 5 (42%, 95% CI 19.3 to 68.1) were attempted by physicians with 4 or more years experience.

Overall there were 57 adverse events resulting in an adverse event rate of 19.7% (95%CI 15.5 to 24.7). There were 31 attempts resulting in cannulation failure (11%, 95% CI 7.7 to 14.8), 20 hematomas (6.9%, 95% CI 4.5 to 10.5) and 6 inadvertent arterial punctures (2%, 95% CI 1.0 to 4.5). Patients with end stage renal disease were more likely to sustain an adverse event. (Table 1) Fewer adverse events were noted among patients who required a vasopressor or inotrope infusion for sustained hypotension while in the ED. Patients undergoing concurrent ACLS experienced more adverse events but this difference was not statistically significant. There were a total of 96 (33%, 95% CI 28.0 to 38.8) obese patients (BMI ≥ 30) and 43 (15%, 95% CI 11.2 to 19.4) anti-coagulated patients (INR≥ 2) but they sustained no significant difference in adverse events. No pneumothoraces or inadvertent placements in the soft tissues occurred. During follow-up, a single patient treated with heparin was noted to develop a small hematoma. However no intervention took place, the central line continued to be used, and heparin was not stopped. Other than this single hematoma no late adverse events were found on follow-up. Overall 52 patients expired (18.4%, 95% CI 14.4 to 23.4) however there was no difference in mortality among those who experienced an adverse event and those who did not.

Table 1.

Demographic and Baseline Clinical Characteristics of Patients Undergoing Central Venous Cannulation in the ED Categorized by Adverse Event

	Adverse Event n=57	No Adverse Event n=232	p –value a
Demographics
Age in Years (Mean, SD)	62.3 (17.3)	61.3 (17.8)	.685 b
Female	26 (45.6)	127 (54.7)	.216
African American	33 (57.9)	135 (58.2)	.968
Comoribidities
BMIc (Median, IQR)	27.1 (22.6–31.8)	27.0 (22.6–32.5)	.924d
Obesec (BMI ≥ 30)	18 (31.6)	78 (33.6)	.641
COPD or Asthma	5 (8.8)	35 (15.2)	.285
History of End Stage Renal Disease	13 (22.8)	20 (8.62)	.003
Clinical Factors
Hypotensive in the ED	34 (59.7)	153 (66.0)	.373
Persistent Hypotension in the ED	13 (22.8)	89 (38.4)	.028
Concurrent ACLS	5 (8.8)	7 (3.0)	.065f
INRe (Median, IQR)	1.3 (1.1–1.7)	1.3 (1.2–1.7)	.599d
INR ≥ 2	8 (14.0)	35 (15.1)	.837
Hospital Mortality	8 (14.0)	44 (19.0)	.385

^aChi Square unless otherwise indicated

^bt-test

^cAvailable for 282 attempts

^dWilcoxon-Mann- Whitney test

^eAvailable for 272 attempts

^fFisher’s Exact Test

There was an association between volume of procedures and adverse event on bivariate testing. (Table 2) There was no statistically significant effect of operator experience according to post-graduate year education or those with a high propensity to use US.

Table 2.

Operator Characteristics

	Adverse Events n=57	Non-Adverse Events n=232	p -value
PGY			.318
Year 1	15 (26.3%)	61 (26.3%)
Year 2	7 (12.3%)	42 (18.1%)
Year 3	19 (33.3%)	52 (22.4%)
Year 4 or greater	16 (28.1%)	77 (33.2%)
Reported No. of Cannulations			.035
Less than 20	5 (8.8%)	29 (12.5%)
20–50	24 (42.1%)	56 (24.1%)
Greater than 50	27 (47.4%)	136 (58.6%)
Not Reported	1 (1.8%)	11 (4.7%)
Likelihood of Utilizing US			.549
In >8 of last 10 attempts	38 (66.7%)	163 (70.3%)
In <8 of last 10 attempts	18 (31.6%)	58 (25.0%)
Not Reported	1 (1.8%)	11 (4.7%)

In our multivariable analysis, history of end stage renal disease and operators who reported intermediate experience (21–50 cannulations) were associated with an increased risk of adverse events. (Table 3) No significant interactions among the variables in our final model were identified. The Hosmer-Lemshow goodness-of-fit test resulted in a p value of .867 denoting good fit. The AUC of the ROC curve was .660 denoting fair model discrimination.

Table 3.

Results of the Multivariable Analysis Identifying Predictors of Ultrasound Guided Internal Jugular Adverse Events^*

Parameter	Parameter Estimate (β)	p-value	Odds Ratio	95% Confidence Interval
History of End Stage Renal Disease	1.26	.002	3.54	(1.59–7.89)
<20 Cannulations vs. >50 Cannulations	.77	.634	.77	(.27–2.24)
20–50 Cannulations vs. >50 Cannulations	.81	.013	2.26	(1.19–4.32)

^*Persistent Hypotension and Concurrent ACLS removed stepwise

Discussion

Several years have passed since the Agency for Healthcare Research and Quality (AHRQ) published its report on “Making Health Care Safer” and promoted US-guided central venous cannulation.¹⁴ In that time two ED based clinical trials comparing US-guided internal jugular cannulation to the anatomical or landmark technique met with enormous success and approval.^{9, 10} In both studies, success rates were over 93% and adverse events were uncommon but the technique did not prove infallible in all patients. In this study we identified patient risk factors associated with an increased risk of adverse events despite the use of US. We found that patients with a history of dialysis and operators with intermediate level of experience were factors associated with adverse events during US-guided internal jugular cannulation. However, among patients with acute adverse events, none required further therapeutic interventions. Only one hematoma was noted on long-term follow up and we found no evidence that any interventions were required to address an adverse event that occurred during ED CVC. Interestingly, coagulopathy, hyperinflation, and obesity, factors long considered to predispose patients to complications, were not associated with adverse events in US-guided lines at the internal jugular site.

Our findings suggest caution in patients with a history of end stage renal disease. Many of these patients require long-term central venous hemodialysis catheters that increase the risk of stenosis and thrombosis.^22–24 Furthermore, patients with strictures may lack any clinical findings such as edema or over engorgement of peripheral veins that would identify high risk cases.^25,12 When visualized under US an engorged central vein due to a distal stricture or thrombus may give the appearance of an easy target.

Obesity, hyperinflation, and coagulopathy are commonly cited risk factors for adverse events related to central venous cannulation.^{14, 16} In our cohort these risk factors were not associated with acute adverse events. Operators may encounter more difficulty with obese or “hyper-inflated” patients at other sites such as the subclavian vein. However, US guidance at the internal jugular minimizes the impact that overlying tissue or vessel depth has on successfully completing the procedure by allowing visualization of the target and needle. Operators can also avoid deep needle passes minimizing the risk of penetration into the chest cavity. BMI alone may be a poor anatomic predictor since it may not reflect neck circumference or anatomical factors that might interfere with the approach to the internal jugular vein. Likewise, coagulopathy was not associated with adverse events. US guidance can reduce the number of needle passes and facilitate cannulation in one puncture likely decreasing the risk of bleeding. Conversely the relative shallowness of the vessel may allow for gentle pressure during unsuccessful attempts decreasing the risk of further bleeding and hematoma formation. While we hesitate to suggest that US guidance would negate the effect of significant anti-coagulation, its application mitigates the risk of adverse events in patients with some degree of coagulopathy.

As prior studies suggest, compared to non-US-guided attempts, novice operators may stand to benefit most from US guidance during cannulation. ^10,14 In our study examining solely US-guided cannulations, operators with intermediate level skills were more likely to experience adverse events compared to novices and those with greater experience. We attribute the effect on novices to the assistance they probably received compared to those with intermediate experience, who were likely performing independently for the first time or who chose to attempt more difficult cases. Our intermediate operators likely represent physicians in a transition phase where they feel they must challenge their skill in order to progress from proficient operator to expert performer.

We hesitate to define “expertise” from our categories of experience since studies in the field of “expert performance” suggest that volume alone may not be sufficient to acquire high skill. Ericsson has demonstrated that the acquisition of expertise depends on individual and environmental factors that combine to propel individuals to acquire superior skills. ²⁶ Familiarity with a particular simple task threatens to arrest the development of necessary cognitive and associated skills that lead to superior performance. Unless operators actively seek improvement, simply accumulating volume may not lead to better performance. ²⁷ Operators who avoid more difficult cases may arrest their development of superior skills.

The participants of early studies comparing US to non-US-guided techniques were conducted among small groups of operators who underwent focused training or among operators with significant experience performing non-US-guided techniques. One group underwent a 2 hour didactic session and defined “experienced” when participants in that study successfully performed only 25 USIJ.¹⁰ In another study with a success rate of 98%, the lead author performed over half of the attempts.⁹ The majority of operators also reported performing volumes exceeding 50 cannulations at the internal jugular site. These groups may represent subsets of physicians who sought additional training or exposure and thereby attained superior performance.

Our setting likely resembled that of institutions seeking to adopt US-guided techniques among physicians with different levels of expertise and interest in US-guided central venous cannulation. Our findings suggest that novice physicians adopting US-guided techniques without rigorous standardized training will likely experience success initially. However, they should remain cautious until they’ve obtained cannulation experience in several scenarios, as early success with US may not indicate the attainment of superior skills with the technology.

Further study into the area of US-guided cannulation must concentrate on validating patient and operator factors associated with acute mechanical adverse events. For example, structured US-based queries in high-risk patients might identify anatomical features that dissuade physicians from attempting cannulation or recognize cases reserved for highly experienced operators. Work with simulators may broaden our understanding of how to distinguish operator proficiency from expertise in the realm of US-guided procedures. Whether simulation exercises can develop the skills required to become an expert operator and decrease acute adverse events in the clinical arena will require future investigation.

Limitations

Our study had several limitations worth discussion. Physicians self-reported their cannulation attempts which might introduce measurement bias. Failed attempts may be under reported thereby minimizing the impact of risk factors. Conversely, successful attempts in patients “perceived” as difficult may be over reported also minimizing the identification of risk factors. However, the proportion of adverse events are similar to those seen in other observational trials both in the ED and in other settings.^28,21 Blinding the outcome assessors would be difficult in this setting. However, future studies could use video recorded attempts to ascertain specific outcomes.

Information regarding operator experience was also self-reported and only measured at one time period. While our data suggested that no single operator crossed training thresholds during the study, some operators may have logged procedures at sites outside the emergency department. This might bias our results regarding operator experience towards the null. Furthermore, the number of performed procedures is not a validated measure for experience and may lead to over or under-reporting. Experience categories based on consensus opinion have identified some thresholds. It is possible that our measure may have over or under emphasized the skills of the operators since technical expertise will not always conform to such categories. However our data does conform to other published US reports where performance thresholds in groups of 20–30 confers some level of proficiency.²⁹

We chose a composite of adverse events for our study outcome, and the interpretation of our results should emphasize the composite of outcomes and not single components. We further chose to study acute mechanical complications but did not consider central line associated blood stream infections in our composite outcome.

Certain adverse events relied on observations that lack a true gold standard. Blinded radiologists evaluated pneumothoraces but arterial punctures and hematomas relied on self-report or observation. Arterial punctures and cannulations tend to make themselves clinically apparent. However there is no standard definition of a clinically relevant hematoma and our cohort had a substantial number of these events. It is likely that this represents a degree of measurement bias given that some hematomas may have gone unreported if the study was not underway. Interestingly, in the one clinical trial that suggested fewer adverse events from US-guided cannulations, over 50% of all adverse events were attributed to hematomas without reporting the clinical consequences.¹⁰ Future studies should develop clinically useful definitions of what constitutes a major hematoma and what constitutes insignificant bleeding.

The statistical diagnostic evaluation of our model revealed good “goodness-of-fit” but only fair discrimination. Our sample, although larger than most prior studies, was relatively small. We were not able to validate our results with an independent sample. However our model did include several frequently cited clinical parameters. The intent of our publication was descriptive and our future efforts will concentrate on validating our results.

Summary statement

US-guided cannulation of the internal jugular vein appeals to many physicians because it increases the efficiency and may decrease adverse events associated with the procedure. The procedure, however, is not infallible. We found that approximately 20% of attempts result in acute mechanical adverse events. We identified both patient (history of end stage renal disease) and physician (intermediate experience level) factors that are associated with acute adverse events.