My contribution and takeaways

  • Wireframing
  • Exploration of various eye tracking tools and software
  • Usability testing using eye tracking
  • Survey designing and interviews
  • Data extraction and visualization

 

Problem 

 

Harmful drug interaction alerts are critical for patient safety but largely ignored by doctors during medication prescribing. Despite efforts to improve alert designs, physicians continue to distrust computerized recommendations.

 

 
 

 

Unsolved challenges

noun_347330_cc.png

Alert fatigue

noun_1037731_cc.png

Poor visual display

noun_152642_cc.png

Inadequate timing and interaction mechanics

noun_1219608_cc (1).png

Lack of trust in given advice

 
 

 

 

    Physicians trust alerts based on

    • Specialization
    • Role in medical hierarchy
    • Demonstrated experience
    • Evidence of understanding the patient's situation
    • Empathy
    • Demonstrated knowledge of evidence from literature
    • Collaborative and Inclusive language
    • Timeliness of the advice

     

    Design directions for Trust-based alerts

     

    Endorsed alerts - Sponsorship of clinical alerts by a specialist , administrator, journal or an individual respected in the local environment.

     

     

    Transparent alerts - Physicians' decisions to comply/ override clinical alerts are made visible in a patient's medical record as an effort towards awareness and self monitoring.

     

     

    Team-sensing alerts - Alerts that would deliver guidance during team discussion, via either high-technology (speech recognition) or low-technology (scribe entering test orders)

     

    Collaborative alerts - Use of inclusive language to support team-building and to mitigate feelings of criticism while encouraging acceptance of advice.

     

     

    Empathy-driven alerts - Incorporating markers of empathy into clinical alerts, such as patient-specific or physician specific information.

     

    Conflict-mitigating alerts - In situations of conflicting advice from specialists, clinical alerts would address this tension (prioritizing different courses of action based on patient's condition)

     

    Agency-laden alerts - Alerts would explicitly leverage individual physicians' notes towards providing recommendations and thus garner an increased level of trust.

     

     
     

     

    Wireframing

     

    Design decisions

    Overlay instead of pop-up

    In order to draw physicians' attention to the warning modal window, an lightbox on an overlay would help them focus on the warning message that the system is trying to communicate. 

    Highlighted button color instead of same colored buttons

    In order to aid physicians avoid dismissing a the warning lightbox without understanding the consequences of their actions, a highlighted button suggests them of the action supposed to be taken.

    Explicit button label instead of Cancel and Okay

    To help physicians avoid dismissing a lightbox without understanding the consequences of their actions, use an explicit button label (Cancel Amiodarone) that states exactly what will happen when the user clicks it.

    High fidelity prototype

    Screen Shot 2018-04-02 at 15.06.28.png
     

     

    Variations of early prototype

     

    Screen Shot 2018-03-12 at 11.45.55.png
    Screen Shot 2018-03-12 at 11.48.20.png
     
    Screen Shot 2018-03-12 at 11.31.56.png
    Screen Shot 2018-03-12 at 11.48.20.png
    Screen Shot 2018-03-12 at 11.32.13.png
     
     
    Screen Shot 2018-03-12 at 11.32.19.png
     
     

     

    Why eye tracking?

     

    • Eye tracking shows physicians' actual eye movements, without any conscious filtering.

    • Helps understand where physicians expect to find certain elements

    • Find out if certain elements are distracting in a negative way

    • Check efficiency of the design in guiding physicians' through the alert message

    • Helps differentiate between new users and experienced users

    • Understand how physicians read - in detail or by scanning

     

    Evaluation using Eye Tracking

     

    Qualitative data collection

     
    TobiiPro_Glasses_2_Eye_Tracker_side_3_1.jpeg
    • Screen recording using Tobii Studio
    • User quotes on each screen design
    • Heatmaps

     

    Quantitative data collection

    • Saccade pathways
    • Behavioral logs  that includes clicks on the main 'theme' of the design and time on task
    • Results from post-task questionnaire using System Usability Scale
    • Eye-tracking data including gaze and fixations on each section on the alerts
    • Preference data about the decision to cancel or continue the order

     

     

    Consolidating heatmaps and resulting saccade pathways

    from 35 physicians

     
     

     

    Outcome

    noun_1387753_cc.png

    Harmful drug interactions reduced by 85%

    noun_1219608_cc.png

    Trusted alerts resulted in better decision making

    Screen Shot 2018-03-12 at 11.01.41.png

    Physicians spending ample time to look at the trust cues before making a decision

     

     

    For details on design process, explorations and iterations, feel free to reach out to me.

     

    Other selected projects

    Tackling crime with predictive policing

    Data visualization, Visual design, Dashboard design

    Protecting the innocent with video redaction

    Interaction design, Product design, Computer vision

    Health insurance made easy

    Visual design, Mobile application design, Information architecture