User-Centered Design in Operational Projects: Illustration with Central Venous Access Device Documentation
TimeThursday, April 152:00pm - 3:00pm EDT
LocationPatient Safety Research and Initiatives
DescriptionAppropriate care of Central Venous Access Devices (CVAD) is critical to reduce risk of line bleeding, thrombosis, and infection. CVAD type and properties determine appropriate care. However, these properties are not adequately captured or easily recognized by users in the Electronic Health Record. We employed User Centered Design to redesign and evaluate CVAD property documentation in EHR.
Central Venous Access Devices (CVAD) are commonly used to give medicine, fluids, blood, or nutrition to a patient by accessing central venous lines near the heart or just inside the heart. CVAD complications such as bleeding, thrombosis, and central line-associated bloodstream infections result in thousands of deaths each year and billions of dollars in added costs to the U.S. healthcare system. In our own health system, we identified safety issues where patients with apheresis ports had thrombosis of their line shortly after insertion due to inappropriate flushing protocols, requiring insertion of new lines and delays in acute care. Further investigation found additional near misses from inadequately documented CVADs including delays in CVAD removal and inappropriate preparation for patient transport while CVADs were accessed.
Appropriate care for CVADs is a complex process that depends on key CVAD properties. For example, high-flow CVADs require heparin locks at different concentrations and different times than low-flow CVADs to avoid thrombosis. To provide appropriate care including flushing protocols, heparin concentrations, locks, and safe removal practices, clinicians must be able to identify if a CVAD is high-flow or low-flow, tunneled or non-tunneled, cuffed or non-cuffed, and if the material is polyurethane or silicone. However, in an internal audit we found these properties were frequently missing from the electronic health record (EHR).
In this study, we adopted UCD to understand the sociotechnical system contributing to poor CVAD documentation quality and design new EHR interfaces for front-line nurses and our vascular access teams. We evaluated the effectiveness of these designs using summative usability tests of experts’ and front-line nurses’ ability to detect key properties and identify appropriate care techniques.
User and Task Analysis and Formative Testing: We first conducted a user and task analysis to understand how front-line and expert users entered and retrieved information for CVAD care. We performed semi-structured interviews with expert providers including IR physicians, surgeons, nurses, and IR techs. We also reviewed institutional policies for CVAD care. Next, we developed a candidate design and performed formative usability testing with a think-aloud protocol with front-line nurses and IR techs in a test version of the EHR. The interface was iteratively adjusted based on observations and participant feedback until no new input was identified in 2 interviews.
Evaluation: We randomly identified 50 CVADs documented pre-implementation and 50 documented post-implementations. Five expert clinicians reviewed the patient chart to identify if line properties were documented correctly and completely. Each chart was reviewed by 2 experts, with one expert reviewer acting as adjudicator.
User and Task Analysis and Formative Testing: A total of 14 clinicians (5 Bedside nurses, 3 Surgical Nurses, 3 IR Techs, 3 Physicians) participated in the user and task analysis. six clinicians (3 physicians and 3 nurses) provided qualitative feedback on the dashboard design. We identified 3 user roles: line placers, line documenters and line care providers. The same individual can act in one or more of these 3 roles, but we found different interactions between these roles in different care settings (e.g. IR vs. Surgery). In our task analysis we identified five key line properties to guide downstream care including flow (high vs low), tunneled (tunneled vs non-tunneled), cuffed (cuffed vs non cuffed), line type (PICC vs Port vs CVL vs Vascath vs Permcath), and material (polyurethane vs silicone). However, information about these key properties was available to each user roles at different times depending on their workflow and level of expertise. Less experienced users, and those documenting lines in floor may not have (1) knowledge of the line the patients presents with, (2) have access to people who place it (3) actual package with line information to help them with documentation of CVAD properties. We discovered that having a forcing function to document all properties made people to select wrong options when they were in doubt. It also thwarted the users to carry on with clinical care without completing the documentation.
Solution: We edited the documentation template, standardized nomenclature to fit mental model and common terms used by frontline users. Key property information previously buried in documentation template were surfaced on titles to allow easy perception and selection of correct lines. To allow for variation in information availability and knowledge gap, we developed a documentation template that allowed an “unknown” option. We also developed dashboard for experts to identify incomplete CVAD documentations.
Evaluation: The documentation of CVAD properties improved significantly (p<0.05) from 56% at baseline to 84% post redesign. Improvements were significant for all key CVAD properties
Table 1. Properties documented in EHR pre and post documentation redesign
EHR Design Tunneled Cuffed Flow Material Overall
Pre-Implementation 66% 57% 60% 43% 56%
Post-Implementation 94% 81% 85% 77% 84%
We employed UCD to improve documentation quality to inform subsequent appropriate care and reduce CVAD complications. We found that expertise and information availability regarding CVAD properties was distributed unevenly between user roles based on their context (e.g. IR vs. Surgery vs. ICU) and experience. We simplified the CVAD documentation interface as much as possible but allowed novice users or those with incomplete information to enter” unknown” for key properties. Dashboards allowed experts to correct incomplete CVAD documentations. UCD led to improved CVAD documentation quality. Future work will (1) Assess front line user ability to extrapolate from documentation to appropriate care (2) Leverage documentation quality for CDS downstream in workflow.