Supplemental Figures

(1)

Supplemental Figures

Supplemental Figure 1. An Example of the GLANCE Interface. The GLANCE interface was designed to contain useful information of glaucoma management in a single window. Demographic information, longitudinal intraocular pressures (IOP), pachymetry measurements, ocular history and medications were included. Additionally, visualizations of IOP, visual field metrics, reported in mean deviation (MD), recent visual fields, as well as the most recent OCT scan and the AI-identified important regions for prediction were shown. Dynamic elements were included, such that hovering over individual data points in the line graph would highlight the corresponding visual field from the same dates. Regions of interest (ROIs) to clinicians are shown in bounded boxes in Supplemental Figure 1A. These ROIs were derived from heatmaps shown in Supplemental Figure 1B.

A.

(2)

B.

(3)

Supplemental Table 1. Survey and Questions sent as part of the GLANCE Usability Study. As part of the GLANCE usability study, demographic data was collected for each user and management

recommendations and attitudes towards the artificial intelligence model’s visual field predictions were collected for each case. At the end of the study, users were asked a standardized questionnaire regarding the usability of the interface in general and given the option to provide additional comments.

Questions

1. What is your current age in years?

2. What is your gender identity?

3. What is your race?

4. What is your ethnicity?

5. Which best identifies your professional role as a clinician?

6. What is your PGY training year?

7. How many years have you been in clinical practice after training?

8. What is your primary subspecialty? - Selected Choice 9. What is your primary subspecialty? - Other - Text

10. Case 1 OD: What is your recommendation for this patient? - Selected Choice 11. Case 1 OD: What is your recommendation for this patient? - Other - Text

12. Case 1 OS: What is your recommendation for this patient's left eye? - Selected Choice 13. Case 1 OS: What is your recommendation for this patient's left eye? - Other - Text

14. Case 1: To what extent to you agree or disagree with the following statements? - I trust the predicted MD enough to incorporate it into my decision-making process when making my clinical recommendation for this patient.

15. Case 1: To what extent to you agree or disagree with the following statements? - The predicted MD provides additional useful information beyond the existing clinical information available to me.

16. Case 1: To what extent to you agree or disagree with the following statements? - I would likely decrease the frequency of visual field testing for this patient if I had predicted MDs available from this algorithm.

17. [optional] Please use this space for additional comments about Case 1

18. Case 2 OD: What is your recommendation for this patient's right eye? - Selected Choice 19. Case 2 OD: What is your recommendation for this patient's right eye? - Other - Text 20. Case 2 OS: What is your recommendation for this patient's left eye? - Selected Choice 21. Case 2 OS: What is your recommendation for this patient's left eye? - Other - Text

42. Case 5 OD: What is your recommendation for this patient's right eye? - Selected Choice 43. Case 5 OD: What is your recommendation for this patient's right eye? - Other - Text

(4)

44. Case 5 OS: What is your recommendation for this patient's left eye? - Selected Choice 45. Case 5 OS: What is your recommendation for this patient's left eye? - Other - Text

50. Case 6 OD: What is your recommendation for this patient's right eye? - Selected Choice 51. Case 6 OD: What is your recommendation for this patient's right eye? - Other - Text

56. Regarding the current prototype of this data visualization system, to what extent do you agree or disagree with the following? - I think that I would like to use this system frequently.

57. Regarding the current prototype of this data visualization system, to what extent do you agree or disagree with the following? - I found the system unnecessarily complex.

58. Regarding the current prototype of this data visualization system, to what extent do you agree or disagree with the following? - I thought the system was easy to use.

59. Regarding the current prototype of this data visualization system, to what extent do you agree or disagree with the following? - I think that I would need the support of a technical person to be able to use this system.

60. Regarding the current prototype of this data visualization system, to what extent do you agree or disagree with the following? - I found the various functions in this system were well integrated.

61. Regarding the current prototype of this data visualization system, to what extent do you agree or disagree with the following? - I thought there was too much inconsistency in this system.

62. Regarding the current prototype of this data visualization system, to what extent do you agree or disagree with the following? - I would imagine that most people would learn to use this system very quickly.

63. Regarding the current prototype of this data visualization system, to what extent do you agree or disagree with the following? - I found the system very cumbersome to use.

64. Regarding the current prototype of this data visualization system, to what extent do you agree or disagree with the following? - I felt very confident using the system.

65. Regarding the current prototype of this data visualization system, to what extent do you agree or disagree with the following? - I needed to learn a lot of things before I could get going with this system.

66. Please provide any comments you feel would be helpful regarding the design of the tool. What features were most helpful? What could be improved? All comments are welcome, and thank you for your time and input.

(5)

Supplemental Figure 2. Mean Percentage of Responses by Age (A) and Role (B) versus Management of Glaucoma Stratified by Severity.

A. Mean % of Responses by Age of User vs Management Stratified by Severity

(6)

B. Mean % of Responses by Role vs Management Stratified by Severity

(7)

Supplemental Figure 3. Overall Mean Likert Scores Assessing Attitude Towards GLANCE. Overall

clinicians found the predicted mean deviation (MD) useful (mean Likert score = 3.42), trustworthy (mean

Likert score = 3.27), but were less inclined to decrease their visual field (VF) testing frequency (mean

Likert score = 2.64) based on the artificial intelligence outputs displayed in the GLANCE interface.

(8)

Supplemental Figure 4. Changes in Mean Likert Scores with Progression through Cases. Mean likert

scores assessing clinician trust in the predicted mean deviation (MD), usefulness of the predicted MD, and

impact on their willingness to decrease visual field testing were all highly variable as users progressed

through the cases. Despite randomization of the cases by severity, clinicians felt less likely to decrease VF

testing frequency as they progressed through the cases.

(9)

Supplemental Figures