Participants
All internal medicine residents (n = 98; postgraduate year (PGY)-1 to PGY-5) were invited to participate in this study between December 2012 and October 2013.
Protocol
At baseline, all consenting participants completed a demographic survey. Participants were then given standardized instructions to place a CVC into the right internal jugular (IJ) vein using ultrasound guidance (SonixTOUCH, BK Ultrasound©) on a simulator (Gen II Ultrasound Central Line Training Model, Blue Phantom™), in a standardized procedure room. The participants were informed that the patient had chronic kidney disease and no peripheral intravenous access. One confederate nursing assistant was in the procedure room and provided assistance as requested by the participants. During the procedure, participants communicated with the patient, whose voice was controlled by researchers in the adjacent control room behind a one-way mirror. The scenario was video recorded using four camera angles, capturing views of the room, procedure site, procedural tray, and ultrasound screen (Fig. 1). Postprocedure, all participants underwent a 30-min semi-structured interview on their strategies for dealing with interruptions.
Interruption
Participants were assigned to two groups in this study. Due to accidental violations in the randomization procedure, the majority of the participants (85%) were not randomized but assigned using unconcealed alternating group assignment. In the control group, participants were interrupted during a task that was felt to be low in complexity: skin cleaning for the insertion site. In the experimental group, participants were interrupted during a more complex task: establishing venous access under direct ultrasound guidance, where the interruption occurred as soon as the venous access needle entered the simulated skin.
At the pre-defined task (i.e., at the time of skin cleaning for the control group and at the time of venous needle skin entry for the experimental group), a 5-s interruption was introduced by a telephone call into the procedure room, whereby the nursing assistant relayed the message on the patient’s high potassium (7.9 mmol/L). An electrocardiogram indicating clinical severity (e.g., peaked T waves and widened QRS) was available if requested by the participant. The nursing assistant was instructed to acknowledge all orders from the participants except for intravenous orders, whereby the participants were reminded that the patient had no intravenous access. As therapy for the hyperkalemia requires intravenous access, it is anticipated that the participants would need to complete the CVC task.
Outcomes
The primary outcomes of interest were (1) overall performance of CVC insertion, (2) time spent on the respective tasks, and (3) number of attempts to establish venous access. Secondary outcomes included results from the thematic analyses of the interviews.
Performance of CVC insertion
Performance of CVC insertion was assessed using a 23-item checklist, modified from a previously published tool with validity evidence, to ensure that the items were applicable to our current task [25, 26].
Items that were executed appropriately were given a score of two, items that were not completed were given a zero, while items that were completed inappropriately or suboptimally were given a score of one. From this checklist, four scores were generated, presented as a percentage:
-
(1)
Overall score: sum of checklist score.
-
(2)
Time 1 score: steps prior to and including cleaning.
-
(3)
Time 2 score: steps after cleaning until venous access establishment.
-
(4)
Time 3 score: remaining steps in the procedure.
All performances were rated by a faculty (IM) with over 10 years of prior experience in rating CVC performances and previously demonstrated high inter-rater reliability using a similar tool [26]. Blinding of the rater to group assignment was not possible as the videos clearly indicate when the interruptions occurred.
Time spent on procedure
Cleaning time was defined as the time taken to clean the insertion site. Time required to access the IJ vein was defined as the time from first needle puncture until successful venous access, as indicated by the removal of the syringe for wire insertion.
Number of attempts
The number of attempts taken to establish venous access using the needle and syringe was recorded. Number of attempts was recorded independently by two researchers (IM and MW). Inter-rater reliability for this measure was high [intraclass correlation coefficient = 0.97, 95% confidence interval (CI) 0.93 to 0.99].
Statistical analyses
Group differences were compared and analyzed in an intention-to-treat basis using standard parametric and non-parametric techniques [27]. Construct validity of the checklist was assessed by comparing performance scores of junior trainees (PGY 1–2) with senior trainees (PGY 3–5): 71.7 ± standard deviation (SD) 22.8 vs. 88.1 ± 5.9%, respectively; p = 0.028. Internal reliability of the checklist was assessed using Cronbach’s alpha (alpha = 0.88).
After testing for the assumption of sphericity (not violated, chi-square (2) = 0.89, p = 0.64, epsilon = 0.96), mixed repeated measure analyses of variances were conducted to assess for group differences on performance scores on the three time points. Partial eta squared values are reported as measures of effect sizes and interpreted as follows: <0.01 = small effect, <0.06 = medium effect, and >0.14 = large effect [28]. Significant interaction between group and time was further explored using the Bonferroni adjustments.
All performances were recorded and time coded with Noldus Recorder and Observer XT, version 11.0 (Noldus Information Technology, Wageningen, the Netherlands). All analyses were performed using SAS 9.3 (SAS Institute Inc., Cary, NC) and PASW Statistics, version 18.0 (PASW, IBM Corporation, Somers NY).
Qualitative data analyses
Interview data was transcribed into NVivo, version 10 (QSR International, Burlington, MA). Thematic content analysis was performed independently by two researchers (IM, JJ) [29]. Assigned codes were reviewed and coded several times to ensure the saturation of themes. Codes were then grouped together based on similarities and linkages to form broader categories. Agreement in coding was high (Kappa = 0.89; 95% CI 0.87 to 0.90) [30]. Disagreement in coding was resolved by consensus.