Successfully Transitioning an Interruptive Alert into a Noninterruptive Alert for Central Line Dressing Changes in the Neonatal Intensive Care Unit

Abstract

Background Interruptive alerts are known to be associated with clinician alert fatigue, and poorly performing alerts should be evaluated for alternative solutions. An interruptive alert to remind clinicians about a required peripherally inserted central catheter (PICC) dressing change within the first 48 hours after placement resulted in 617 firings in a 6-month period with only 11 (1.7%) actions taken from the alert.

Objectives This study aimed to enhance a poorly functioning interruptive alert by converting it to a noninterruptive alert aiming to improve compliance with the institutional PICC dressing change protocol. The primary outcome was to measure the percentage of initial PICC dressing changes that occurred beyond the recommended 48-hour timeframe after PICC placement. Secondary outcomes included measuring the time to first dressing change and, qualitatively, if this solution could replace the manual process of maintaining a physical list of patients.

Methods A clinical informatics team met with stakeholders to evaluate the clinical workflow and identified an additional need to track which patients qualified for dressing changes. A noninterruptive patient column clinical decision support (CDS) tool was created to replace an interruptive alert. A pre–postintervention mixed-methods cohort study was conducted between January 2022 and November 2022.

Results The number of patients with overdue PICC dressing changes decreased from 21.9% (40/183) to 7.8% (10/128) of eligible patients (p < 0.001), and mean time to first PICC dressing changes also significantly decreased from 40.8 to 30.7 hours (p = 0.02). There was a universal adoption of the CDS tool, and clinicians no longer used the manual patient list.

Conclusion While previous studies have reported that noninterruptive CDS may not be as effective as interruptive CDS, this case report demonstrates that developing a population-based CDS in the patient list column that provides an additional desired functionality to clinicians may result in improved adoption of CDS.

Keywords

clinical decision support - alert fatigue - noninterruptive alerts - neonatology - intensive care

Background and Significance

The Agency for Healthcare Research and Quality funded a national effort to prevent central line-associated bloodstream infections (CLABSI) due to the profound impact on patient outcomes and costs and the generally recognized preventability of the condition.[1] It is well-recognized that hospital compliance with central line bundles and protocols is associated with a reduction in CLABSI rates.[2] [3] Thus, hospitals frequently monitor for increased infection rates and may implement interruptive alerts, dashboards, or checklists into the electronic health record to remind clinicians to follow locally created CLABSI protocols in an attempt to prevent this severe adverse event.[4] [5] [6]

Interruptive alerts are a major contributor to alert fatigue causing clinicians to override approximately 90% of alerts, which may lead to information overload and adverse events.[1] [2] Experts recommend that hospitals have a clinical decision support (CDS) review and governance committee that regularly reviews poorly performing alerts and considers redesigning interruptive alerts with alternative forms of CDS.[7] [8] In a quality improvement project, investigators reported that decreasing the number of alerts clinicians were exposed to resulted in improved interaction with other alerts.[9] Thus, evaluating poorly performing interruptive alerts and assessing if they can be transitioned into alternative forms of CDS may result in improved compliance with the primary measure and other alerts.

Context

At the University of Iowa Neonatal Intensive Care Unit (NICU) a quaternary care 88-bed unit, a peripherally inserted central catheter (PICC) protocol requires that licensed advanced practice providers (APPs) who are trained in placing PICCs perform a dressing change of the PICCs within 48 hours of placement to evaluate for signs of infection or needed readjustment of the line after repeat chest radiograph. Oozing of blood from the insertion site is common during the 24 hours after insertion and increases the risk of infection or line migration, thus, a dressing change within 48 hours is recommended.[10] However, due to team-based care and alternating schedules of clinicians, a different APP than the one who placed the PICC may be responsible for the dressing change. Historically, the list of patients who needed dressing changes was kept on a physical whiteboard in a clinical workroom for the clinicians to review every day. This solution was prone to omissions after a new PICC was placed. Thus, a request was placed to create an interruptive alert to remind providers to perform a dressing change within 48 hours whenever a new PICC was documented on a patient ([Fig. 1]). However, there was negative feedback on the interruptive alert malfunctioning and causing interruptions to clinical workflow.

Fig. 1 Example of the pop-up interruptive alert reminding clinicians that they needed to document a PICC dressing change in the next 24 hours to prevent the patient from becoming overdue for their first dressing change. For team members unable to changing PICC line dressings they were expected to consult the charge nurse practitioner. Epic 2022. PICC, peripherally inserted central catheter.

Problem Statement

Clinicians had difficulty identifying patients who qualified for their first PICC dressing change based on the hospital CLABSI bundle, and the interruptive alert that was implemented caused clinician dissatisfaction and did not achieve the desired performance.

Methods

The objective of this implementation case study was to enhance a poorly functioning interruptive alert by converting it to a noninterruptive alert aiming to improve compliance with the institutional PICC dressing change protocol. The primary outcome was to measure the percentage of initial PICC dressing changes that occurred beyond the recommended 48-hour timeframe after PICC placement. CLABSI rates associated with these changes were not monitored, as the dressing change is only one piece of the CLABSI bundle, and this alert change and study time span would not be expected to provide enough power to affect CLABSI rates. Secondary outcomes included measuring the time to first dressing change and qualitatively monitoring if this solution would replace the manual process of writing patients' names on the whiteboard.

Using the systems engineering initiative for patient safety (SEIPS) 101 model,[11] the investigators described the work system, tools, and interactions to see if the primary objective of improved compliance with the protocol would be achieved with the proposed tool ([Supplementary Table S1] [available in the online version only]). First, firing statistics were reviewed, and comments on the interruptive alert describing why the alert was inaccurate were evaluated. Then, clinical informatics team members met with clinical stakeholders to understand the clinical work system and review why the current interruptive alert was ineffective. The original workflow included a member of the nurse practitioner team trained in placing PICCs who would place a PICC successfully and then add the patient to the whiteboard for the team to track for 24 to 48 hours until the initial dressing change occurred.

Based on the need to both identify and track which patients have had a PICC placed recently and if the initial dressing change was performed within the protocol time frame of 48 hours, the development team created a new column in the patient list ([Fig. 2]). The patient list is the main list of all patients a provider cares for that day. Clinicians use it to see the unit census and open patients' charts. In the NICU, the institution had not previously provided decision support through the patient list. The new column in the patient list was built with rules using discrete data that assessed if a new PICC was placed based on the lines, dressing, and airway documentation (LDA). Another rule then evaluates if the first dressing change was documented in the nursing flowsheet discrete data. The interruptive alert inaccurately fired with failed PICC placement attempts and patients who transferred in with PICCs that did not require dressing changes. To decrease the inaccuracy of the alert, we updated the logic in the noninterruptive alert to only look to LDA placement instead of the PICC order, and bedside nurses were educated to complete the PICC dressing change documentation upon admission of transferred patients. A red, yellow, and green light indicator quickly communicates to clinicians which patients need to be prioritized for the day ([Fig. 2]). Once the patient has their first dressing change documented in the flowsheets, the column becomes blank.

Fig. 2 Clinical decision support in the patient list column with green displaying a new PICC placed in the past 24 hours, yellow displaying the dressing is due to be changed soon (24–48 hours since placement), and red if the PICC dressing change is overdue (>48 hours since placement). Epic 2022. PICC, peripherally inserted central catheter.

This simple design was chosen because a new patient column could be added to any inpatient list to target the appropriate clinicians who are skilled in changing PICC dressings, and team members who were not responsible for this task did not have to add this CDS to their patient list. This list allowed clinicians to review the CDS whenever it fits into their clinical workflow for the day. Education for this new CDS tool was provided via email and was directly communicated by clinician champions to the providers trained in performing PICC dressing changes (n = 12).

Data Collection

A pre-postintervention mixed-methods cohort study was conducted between January 2022 and November 2022. A preimplementation analysis was performed using Structured Query Language (SQL) report including the date and time a new PICC was documented and the date and time the first dressing change was documented (January 2022–June 2022). The number of hours to first dressing change was calculated by this report. The same postimplementation report was analyzed 4 months after implementation of the noninterruptive CDS (July 2022–November 2022) to evaluate if the new intervention was successful. Statistical methods used included t-test and chi-square test for continuous and categorical variables, respectively, since the data were normally distributed. We plotted a Kaplan–Meier survival curve of time to first dressing change across the pre- and postimplementation groups and compared the equality of the survivor function across groups using STATA version 17.0 (StataCorp, College Station, Texas, United States). Patients with time to first dressing change that was greater than 100 hours were censored for the Kaplan–Meier curve data visualization.

Results

Interruptive Alert Analysis

Baseline analysis of the interruptive alert ([Fig. 1]) to remind clinicians that a PICC dressing was due within the next 24 hours fired 617 times during a 6-month period (January 2022–July 2022) with only 11 firings (1.7%) that resulted in any action being taken. Users chose the 10-minute snooze button 282 times (46% of total firings). Upon further discussion with clinical stakeholders, it was found that this alert was not effective because it frequently occurred during inappropriate times in their workflow when they were not able to perform the task which led to dependence on the manual process of a list on the whiteboard. Additionally, it did not provide the necessary functionality of a real-time list of patients who qualified for dressing changes within the next 48 hours.

Noninterruptive Alert Analysis

In July 2022, the noninterruptive solution was available to be added to the patient list of clinicians who signed into the neonatal critical care context, and the interruptive pop-up alert was retired. All the clinicians on the PICC placement team (n = 12) adopted the patient list column. This optional column allowed team members who did not have the expertise in changing a PICC line dressing to avoid this CDS alert. Team members collaborating on completing the required dressing changes for the day could view a real-time list of which patients to prioritize.

After the implementation of the patient column, the number of patients with overdue PICC dressing changes decreased from 21.9% (40/183) of eligible patients to 7.8% (10/128) of eligible patients (p < 0.001). The mean time to perform the initial dressing change decreased from 40.8 hours (95% CI 34.1–47.6) to 30.7 hours (95% CI 28.3–33.1; p = 0.02). The Kaplan–Meier survival analysis showed that the time to first dressing change improved with our noninterruptive intervention ([Fig. 3]; p = 0.02). The interruptive alert group had eight (4%) patients who did not have a documented first dressing change within 100 hours and the noninterruptive group had no patients with delayed recognition of first dressing change >100 hours. After the universal adoption of this method by the key stakeholders, the manual practice of writing patients on the whiteboard who needed PICC line dressing changes was retired.

Fig. 3 Kaplan–Meier curves showing time to first dressing change within the desired 48-hour protocol which show that the noninterruptive alert led to a timelier dressing change (p = 0.01). Eight patients in the interruptive alert cohort were censored from this graph due to first dressing change occurring >100 hours after PICC placement. PICC, peripherally inserted central catheter.

Discussion

By using the SEIPS 101 method of first understanding the work system and then evaluating the proposed tools and how they interact with the people and environment, a noninterruptive CDS solution that easily fits into the clinical workflow was designed, implemented, and achieved the desired primary and secondary outcomes. The noninterruptive CDS significantly decreased the number of events that violated the initial dressing change protocol, decreased the time to first dressing change, and subjectively allowed providers to monitor the dressing change workload efficiently based on user feedback. Moving the CDS from a patient-based pop-up interruptive alert to a population-based patient list column allowed for quick prioritization of which patients needed attention and could be targeted only to the team members with the skills to perform the task. Due to the small size of the team specialized to perform this procedure, directed education was easy to perform which likely improved the success of this CDS. In addition, patient-list CDS may provide more frequent reminders of eligible patients than traditional interruptive pop-up alerts, since every time a clinician opens their patient list, they have an opportunity to review the CDS.

While others have reported using electronic dashboards or checklists to help with the maintenance of invasive catheters, we have been unable to find reports of using patient list columns.[4] [6] Previous literature states that some forms of noninterruptive CDS may not be as effective as interruptive CDS.[12] [13] This report demonstrates that developing CDS that provides additional functionality to clinicians with a real-time ability to track the workload for the day improved the adoption and awareness of this new CDS tool. Reports of usability testing of noninterruptive alerts have shown that there is less visibility with noninterruptive CDS.[14] This highlights the importance that noninterruptive CDS should be monitored after going live to measure if the desired outcomes are achieved. Noninterruptive CDS has been shown to improve patient safety workflows in the perioperative environment during barcode scanning, however, few case studies of successful noninterruptive CDS have been reported.[15] Factors contributing to the success of this CDS may be due to the simple population-based design and the colors that highlight when patients are overdue for a dressing change.

Limitations of this study include a single-center study where this intervention was only tested on a small patient population in the NICU over a relatively short time frame. In addition, we were unable to completely isolate the testing of an interruptive alert to a noninterruptive alert as this study did have several other important intervention changes such as transitioning to a population-based alert, implementing a colored coded time-sensitive feature, and targeted training and education to specific clinicians. A significant limitation to our design is that it does not account for color-blind clinicians which may be improved upon by incorporating icons into future iterations. Additional limitations include that the study did not track the impact of this implementation on the CLABSI rate and an inability to assess how many times a day the clinician actively reviewed the patient list CDS. Since we only implemented this solution in one unit in our hospital that used this protocol, we performed a unit-wide universal implementation. However, implementing a similar patient column CDS solution hospital-wide may benefit from a step-wedge implementation to monitor the effectiveness of the alert in different units in the hospital.

Clinical Relevance Statement

This work displays an innovative CDS solution to improve compliance with a unit protocol and CLABSI bundle and helps providers prioritize and manage their workflow tasks for the day. This noninterruptive CDS solution is a simple example of how interruptive alerts that distract and annoy clinicians can be transformed into alternative forms of CDS that may be more effective at achieving the desired clinical outcome.

Multiple Choice Questions

1. Is noninterruptive CDS more effective than interruptive CDS?

   • True

   • False

   Correct Answer: The correct answer is option b. False. After going live noninterruptive CDS must be monitored to see if it can achieve the desired clinical outcome(s). Usability testing has shown that noninterruptive CDS generally has less visibility than interruptive CDS. The decision to use noninterruptive CDS should depend on the clinical context and clinical workflow.

2. Interruptive alerts have been associated with which of the following:

   • Alert fatigue

   • Burnout

   • Preventing adverse events

   • All of the above

   Correct Answer: The correct answer is option d. All of the above. Interruptive alerts may be developed as safety checks and reminders of recommended best practices for patients. However, with the overwhelming number of interruptive alerts that have been implemented into the electronic health record (EHR), they have also been associated with alert fatigue and clinician burnout. Therefore, all hospitals should have a CDS review team to monitor the overall number of alerts firing in the system and to try to reduce ineffective or unnecessary alerts, so clinicians are only exposed to clinically relevant alerts.

Conflict of Interest

None declared.

Protection of Human Subjects

The University of Iowa IRB determined this project to be non–human subject research.

• References

• 1 Agency for Healthcare Research and Quality. Eliminating CLABSI. A National Patient Safety Imperative: Final Report Companion Guide. January 2013 . Accessed August, 2023 at: https://www.ahrq.gov/hai/cusp/clabsi-final-companion/index.html
• 2 Gavin NC, Webster J, Chan RJ, Rickard CM. Frequency of dressing changes for central venous access devices on catheter-related infections. Cochrane Database Syst Rev 2016; 2 (02) CD009213
• 3 Tripathi S, McGarvey J, Lee K. et al. Compliance with central line maintenance bundle and infection rates. Pediatrics 2023; 152 (03) e2022059688
• 4 Pageler NM, Longhurst CA, Wood M. et al. Use of electronic medical record-enhanced checklist and electronic dashboard to decrease CLABSIs. Pediatrics 2014; 133 (03) e738-e746
• 5 Rabbani N, Pageler NM, Hoffman JM, Longhurst C, Sharek PJ. Association between electronic health record implementations and hospital-acquired conditions in pediatric hospitals. Appl Clin Inform 2023; 14 (03) 521-527
• 6 Davis CL, Bjoring M, Hursh J. et al. The intensive care unit bundle board: a novel real-time data visualization tool to improve maintenance care for invasive catheters. Appl Clin Inform 2023; 14 (05) 892-902
• 7 Chaparro JD, Beus JM, Dziorny AC. et al. Clinical decision support stewardship: Best practices and techniques to monitor and improve interruptive alerts. Appl Clin Inform 2022; 13 (03) 560-568
• 8 McGreevey III JD, Mallozzi CP, Perkins RM, Shelov E, Schreiber R. Reducing alert burden in electronic health records: state of the art recommendations from four health systems. Appl Clin Inform 2020; 11 (01) 1-12
• 9 Simpao AF, Ahumada LM, Desai BR. et al. Optimization of drug-drug interaction alert rules in a pediatric hospital's electronic health record system using a visual analytics dashboard. J Am Med Inform Assoc 2015; 22 (02) 361-369
• 10 Pettit J, Wyckoff MM. Peripherally Inserted Central Catheters. Guideline for Practice, 2nd ed. National Association of Neonatal Nurses;; 2007: 44
• 11 Holden RJ, Carayon P. SEIPS 101 and seven simple SEIPS tools. BMJ Qual Saf 2021; 30 (11) 901-910
• 12 Trinkley KE, Blakeslee WW, Matlock DD. et al. Clinician preferences for computerised clinical decision support for medications in primary care: a focus group study. BMJ Health Care Inform 2019;26(01):0
• 13 Pevnick JM, Li X, Grein J, Bell DS, Silka P. A retrospective analysis of interruptive versus non-interruptive clinical decision support for identification of patients needing contact isolation. Appl Clin Inform 2013; 4 (04) 569-582
• 14 Blecker S, Pandya R, Stork S. et al. Interruptive versus noninterruptive clinical decision support: Usability study. JMIR Hum Factors 2019; 6 (02) e12469
• 15 Steitz BD, Li G, Wright A, Dunworth B, Freundlich RE, Wanderer JP. Non-interruptive clinical decision support to improve perioperative electronic positive patient identification. J Med Syst 2022; 46 (03) 15

Publication History

Received: 25 January 2024

Accepted: 18 August 2024

Accepted Manuscript online:
20 August 2024

Article published online:
13 November 2024