Page 402 -

392 CHAPTER 13 Measuring the human

LAB-IN-A-BOX—CONT'D

physicians see EMRs as bringing changes in documentation, communication,
and work processes, along with concerns about data quality (Embi et al.,
2004), understanding the dynamics of how these records impact care is
more challenging. A 2016 literature review found that although some studies
found that although EMR use involved a range of both positive and negative
communication behaviors, there was no conclusive evidence of any negative
impact on patient perceptions of satisfaction or communication with physicians
(Alkureishi et al., 2016).
Although these results suggest that in-depth studies of the use of EMRs
during patient visits are needed to understand specific behaviors and to separate
negative from positive impacts, conducting such studies presents several
challenges. Lab-based simulations are likely too artificial, lacking the
open-ended challenges associated with medical practice. Some researchers have
resorted to video and audio recordings providing data capture from multiple
perspectives (Asan and Montague, 2014). This approach is informative, but
limited, as these captures might be able to identify where users are looking and
how they are interacting, but details of interactions with the EMR will not be
recorded, leaving researchers with the challenge of inferring how the details of
the computer use might impact communication with patients.
Noting these difficulties, Nadir Weibel and colleagues developed a data
and analysis infrastructure known as LAB-IN-A-BOX, designed to capture
multiple streams of data detailing the dynamics of interactions between the
physician, the patient, and the computer during medical visits. LAB-IN-A-BOX
combines directional audio through a microphone array; eye tracking; full-room
video, screen capture, mouse movements, mouse clicks, and other computer
interactions through Techsmith Morae usability software (https://www.
techsmith.com/morae.html); and Kinect for Windows to measure orientation
of the user's body (Weibel et al., 2014). Realizing that using this device in
physician examination rooms would require a great deal of flexibility in
transportation and installation, Weibel and colleagues configured a hard plastic
rolling case to hold all of the equipment, wiring, and connectors, enabling setup
and data collection in 10 minutes or less (Weibel et al., 2014) (Figure 13.8).
To address the challenge of analyzing the various data streams, Weibel
and colleagues started with a synchronization algorithm that aligns audio and
video components. Kinect data is segmented to differentiate (when possible)
between the clinician, the patient, and objects in the room such as chairs—
all problems that would not be faced in an idealized lab environment with
only one participant and no furniture. This data is then further processed to
determine where the physician is looking at any given time, and to identify
any gestures. Directional audio is processed to distinguish physician speech
from patient speech. Mouse, keyboard, and other computer activities also are