Characterizing and Visualizing Display and Task Fragmentation in the Electronic Health Record: Mixed Methods Design

Abstract

Background: The complexity of health care data and workflow presents challenges to the study of usability in electronic health records (EHRs). Display fragmentation refers to the distribution of relevant data across different screens or otherwise far apart, requiring complex navigation for the user's workflow. Task and information fragmentation also contribute to cognitive burden.

Objective: This study aims to define and analyze some of the main sources of fragmentation in EHR user interfaces (UIs); discuss relevant theoretical, historical, and practical considerations; and use granular microanalytic methods and visualization techniques to help us understand the nature of fragmentation and opportunities for EHR optimization or redesign.

Methods: Sunburst visualizations capture the EHR navigation structure, showing levels and sublevels of the navigation tree, allowing calculation of a new measure, the Display Fragmentation Index. Time belt visualizations present the sequences of subtasks and allow calculation of proportion per instance, a measure that quantifies task fragmentation. These measures can be used separately or in conjunction to compare EHRs as well as tasks and subtasks in workflows and identify opportunities for reductions in steps and fragmentation. We present an example use of the methods for comparison of 2 different EHR interfaces (commercial and composable) in which subjects apprehend the same patient case.

Results: Screen transitions were substantially reduced for the composable interface (from 43 to 14), whereas clicks (including scrolling) remained similar.

Conclusions: These methods can aid in our understanding of UI needs under complex conditions and tasks to optimize EHR workflows and redesign.

Keywords: data visualization; electronic health record; electronic medical record; information technology; medical informatics; user computer interface.

Figure 1

An example of sunburst chart data in Excel describing the system architecture (A) and the resulting sunburst diagram (B).

Figure 2

Interactive sunburst highlighting one pathway (in green) from a root screen (Provider View) to a specific element (Legionella Antigen Urine lab results). The traced pathway is also described in the linear flow above the sunburst diagram. This diagram shows how 9 different screens must be navigated to access the desired element from the main screen.

Figure 3

Display Fragmentation Index element calculations. DFI: Display Fragmentation Index.

Figure 4

An example of time belt visualization for 5 patient cases in preoperative care at a large tertiary care hospital. One single horizontal belt or row represents 1 patient case. The length of the belt indicates the case duration in seconds. Each belt comprised a sequence of tasks performed by the nurse and represented as color-coded segments. For example, Allergies refers to the task of checking allergies.

Figure 5

Sunburst diagram representing display fragmentation of clinical data in a conventional, commercial electronic health record. Elements colored in black are those relevant for handling the clinical problem (general review of patient information). EHR: electronic health record; UI: user interface.

Figure 6

Sunburst diagram representing display fragmentation of clinical data in a second conventional, commercial electronic health record. Elements colored in black are those relevant for handling the clinical problem (general review of patient information). EHR: electronic health record.

Figure 7

Conventional electronic health record (EHR) system user interface (UI).

Figure 8

Time belt visualization of clinical task using commercial electronic health record. Note that the labels for the tasks have been abbreviated for readability. Each item represents a task, primarily searching and reviewing tasks. For example, “X-Ray” is short for “Reviewing X-Ray,” a task the clinician completed.

Figure 9

Experimental system screen with user placement of data elements for the same case as in Figure 8.

Figure 10

Time belt visualization of clinical tasks using experimental electronic health record (EHR) system.

Figure 11

Color of each cell or segment represents the task or screen the user was completing in sequential order. The time belt visualization shows the time taken for each task, whereas the Excel data scheme shows the different screen or part of the system needed for each task.

Figure 12

An example of proportion per instance calculated on the basis of the time belt from Figure 11. Proportion per instance for certain subtasks have been calculated in the small table to the right as examples. PPI: proportion per instance.