Background
Data integration and visualization software
Overview of project phases
Methods
Study design
Setting
Software
Scenarios and tasks
Scenario number | Main events or interventions | Number of parameters | Key data features |
---|---|---|---|
1 | - 2 episodes of hypotension - 1 cardiac arrest - 1 initiation onto extracorporeal membrane oxygenation | 32 total 24 active | - physiological monitoring - infusion pump data - temperature data - laboratory data |
2 | - 1 increased erroneous, medical infusion (dopamine) - 1 intervention (inhaled nitric oxide therapy) | 34 total 28 active | - physiological monitoring - infusion pump data - laboratory data |
3 | - 1 attempt at bedside chest closure - 1 cardiac arrest | 46 total 36 active | - physiological monitoring - infusion pump data - ventilator data - three oxygen saturation parameters - laboratory data |
Participants
Procedure
Data analysis
Scoring task completion and usability error definition: Use error rating
Normative Use Error Rating | Numerical Use Error Rating | Definition | |
---|---|---|---|
Pass | 2 | User completed task with no hint, clarification, or reminder | |
Help | 1 | User completed task with one hint | |
Fail | 0 | User did not complete task despite several hints |
Usability issue severity level
Results
Participants
Physicians | Nurses | Respiratory Therapists | Global Proportion | ||
---|---|---|---|---|---|
Total Number | 7 | 8 | 7 | 22 | |
Gender, % (n) | Male | 14 (n = 1) | - | 14 (n = 1) | 9 (n = 2) |
Female | 86 (n = 6) | 100 (n = 8) | 86 (n = 6) | 91 (n = 20) | |
ICU Experience, % (n) | <1 year | 43 (n = 3) | 13 (n = 1) | 29 (n = 2) | 27 (n = 6) |
1–3 years | 29 (n = 2) | 25 (n = 2) | - | 18 (n = 4) | |
4–10 years | 29 (n = 2) | 25 (n = 2) | 57 (n = 4) | 36 (n = 8) | |
>10 years | - | 38 (n = 3) | 14 (n = 1) | 18 (n = 4) | |
ICU Shifts/Week, % (n) | 1–2 times/week | - | 25 (n = 2) | 29 (n = 2) | 18 (n = 4) |
3–4 times/week | 29 (n = 2) | 75 (n = 6) | 71 (n = 5) | 59 (n = 13) | |
>4 times/week | 71 (n = 5) | - | - | 23 (n = 5) | |
ICU Specialization, % (n) | CCCUa
| 29 (n = 2) | 63 (n = 5) | - | 32 (n = 7) |
PICUb
| 29 (n = 2) | 38 (n = 3) | - | 23 (n = 5) | |
PICU/CCCU | 43 (n = 3) | - | 100 (n = 7) | 45 (n = 10) | |
Previous Training with Software, % (n) | Yes | 14 (n = 1) | 50 (n = 4) | 14 (n = 1) | 27 (n = 6) |
No | 86 (n = 6) | 50 (n = 4) | 86 (n = 6) | 73 (n = 16) | |
Software Use/Shift, % (n) | Several times/shift | 29 (n = 2) | - | - | 9 (n = 2) |
Once/shift | 14 (n = 1) | 13 (n = 1) | - | 9 (n = 2) | |
Rarely during a shift | 43 (n = 3) | - | - | 14 (n = 3) | |
Never | 14 (n = 1) | 88 (n = 7) | 100 (n = 7) | 68 (n = 15) | |
Awareness of Software, % (n) | Yes | 71 (n = 5) | 75 (n = 6) | 43 (n = 3) | 64 (n = 14) |
No | 29 (n = 2) | 25 (n = 2) | 57 (n = 4) | 36 (n = 8) |
Interrater reliability
Software strengths (aid to task completion) and usability issues (hindrance to task completion)
Overview
General Functions | Tasks Tested for Each Function | Error Severity Level | Average Use Error Rating by Task and by Clinician Type | Average Use Error Rating by Task | ||
---|---|---|---|---|---|---|
Physicians (n = 7) | Nurses (n = 8) | Respiratory Therapists (n = 7) | ||||
Tracking: Orientation (4 tasks) | 1. Locating patient file | High | P (2.0) | P (2.0) | P (2.0) | P (2.0) |
2. Identifying a value for a specific physiological variable | High | P (1.8) | P (1.8) | P (1.5) | P (1.7) | |
3. Estimating duration of event by identifying two time points | High | H (1.4) | P (2.0) | P (2.0) | P (1.8) | |
4. Manipulating time scale | High | H (1.0) | F (0.4) | H (0.6) | H (0.6) | |
Function Use Error Rating by Clinician Type | P (1.5) | H (1.5) | P (1.5) | P (1.5) | ||
Trajectory: Relationships between Parameters (10 tasks) | 5. Comparing trends for two specific parameters | High | H (1.4) | P (1.6) | P (1.5) | H (1.5) |
6. Comparing different patient physiological states | High | H (1.3) | H (1.4) | H (1.2) | H (1.3) | |
7. Identifying values for two specific parameters at an event | High | H (1.4) | H (1.1) | H (0.6) | H (1.0) | |
8. Identifying vital signs (group of parameters) prior to an event | High | H (0.7) | F (0.4) | H (1.3) | H (0.8) | |
9. Viewing trend of three redundant overlapping parameters | High | H (1.3) | H (1.4) | H (0.7) | H (1.1) | |
10. Viewing infusion medication data | High | P (1.8) | H (1.3) | P (2.0) | P (1.7) | |
11. Comparing infusion medications with vital signs | High | P (1.9) | P (1.7) | P (1.6) | P (1.7) | |
12. Detecting change in infusion medication rate over time | High | H (1.4) | F (0.4) | H (0.5) | H (0.8) | |
13. Viewing ventilator data | High | P (2.0) | P (1.6) | P (1.6) | P (1.7) | |
14. Viewing laboratory data | High | H (1.0) | P (1.8) | P (1.7) | H (1.5) | |
Function Use Error Rating by Clinician Type | H (1.4) | H (1.3) | H (1.3) | H (1.3) | ||
Triggering: Automated Integration (3 tasks) | 15. Viewing target ranges using shading (semi-automatic aid) | Moderate | F (0.4) | H (0.6) | F (0.4) | F (0.5) |
16. Sparkline (automatic trend line for one variable) | Minor | F (0.4) | H (0.8) | F (0.0) | F (0.6) | |
17. IDO2 indicator (automatic computation using 16 parameters) | High | F (0.4) | H (0.5) | F (0.3) | F (0.4) | |
Function Use Error Rating by Clinician Type | F (0.4) | H (0.6) | F (0.2) | F (0.4) | ||
Other Functions (3 tasks) | 18. Finding notes | High | H (1.1) | P (1.9) | H (1.4) | H (1.5) |
19. Modifying/adding note | Moderate | H (0.9) | H (1.3) | P (1.5) | H (1.2) | |
20. Setting targets | Moderate | P (1.9) | P (1.9) | P (2.0) | P (1.9) | |
Function Use Error Rating by Clinician Type | H (1.3) | P (1.7) | P (1.6) | P (1.5) | ||
All functions | Global Function Use Error Rating, for All Functions by Clinician Type and for All Clinicians | H (1.3) | H (1.3) | H (1.2) | H (1.2) |
Tracking function
Trajectory function
Triggering function
Other functions: Charting
Summary of results
Discussion
Transforming numerical point data to long-term, time-scaled visualizations
Integrating data trends: Visual pattern overload
Pre-defined parametric grouping
Scaling according to the nature of the parameters
Data reduction using algorithms
Novel visualizations
Data trustworthiness
Usability testing with diverse clinician groups
Proposed iteration and improvements
Improvement | Rationale | Suggestions to Achieve Improvement |
---|---|---|
Reduce redundant data streams. | Removal of redundant data is required to allow clinicians to efficiently and easily abstract, trend, and interact with the data. | Ensure preprocessing mass volumes of continuous real-time data. For example, employ algorithms that corrects etCO2trends using pCO2 blood gas values. |
Provide user awareness. | User-aware applications that dynamically adjust the data display mode based on the user context can ensure that adequate and relevant data needs are being displayed and enhance clinicians’ efficiency and efficacy in extracting meaningful information. | Provide customized view of patient data tailored to the clinician’s needs. For example, if a respiratory therapist is detected as the user, the system would display etCO2 with pCO2 to help respiratory therapist know if s/he should trust the etCO2 continuous trend data. |
Reduce clinician cognitive demand in interacting with the visual displays. | Ensuring that components that are important for decision-making are represented in the display in a perceptually similar manner as to improve the clinician’s decision-making accuracy and efficiency. | Present the components that are important for decision-making as an integrated object and/or by presenting them close together spatially or temporally. |
Mandate integration of data integration and visualization software with existing medical record systems. | Integration of data integration and visualization software with medical record systems to provide a single source of patient data which facilitates data synchronization and may reduce use errors. | Technology procurement policies should require incoming data platforms to freely exchange data and information with existing clinical information systems. |
Provide easy time navigation. | A critical function of the interface is enabling the user to rapidly select the time frame of continuous data, relative to the patient’s stay in the ICU. | Provide interface controls which support both exploratory data navigation across time and specific user defined timeframes. |
Ensure interface is flexible to different types of users and levels of expertise. | Functions which are learned should provide shortcuts for accelerated performance. | Provide layered function description and interface shortcuts. |
Ensure software responsiveness. | Additional data streams and access to denser data visualizations may slow down system performance and diminish user satisfaction and decision-making quality. | Ensure new data streams are compressed or back-end processing is sufficient to maintain adequate responsiveness. |