Prolonged Exposure Collective Sensing System (PECSS)
A platform to administer and visualize patient's progress during PTSD therapy
Outcome
Ongoing Project
Role
UX Research,
UX Design,
Android Development
Team
Dr. Rosa Arriaga, Catherine Deeter, Hayley Evans, Varnit Jain, Adam Hayward, Dr. Thomas Ploetz, Peter Presti, Dr. Andrew Sherrill, Marcus Wilder
Tools
Qualtrics, Sketch, Android Studio, Java, JavaScript, Python
National Science Foundation
Principal Investigator: Dr. Rosa Arriaga
Graduate Research Assistantship supported by NSF 1915504 since August 2019.
Timeline (Aug '19 - Present)

Research
6 months
Generative research with clinicians and veterans
Design
2 month
Designing the patient-facing application interface
Android Development
10 months
Completed the development of the first version of the HIPAA-compliant PECSS application
Context
Post Traumatic Stress Disorder (PTSD) is a chronic condition marked by considerable distress and dysfunction​. Among the many pharmacological and psychotherapy approaches that have been used to treat PTSD, Prolonged Exposure (PE) therapy ​has the best evidence for therapeutic efficacy.
"However, the delivery of PE Therapy has been constrained by data collected from patient self-report and clinician intuition. This data is subjective and narrow, functioning as an ever-present obstacle in the practice, training, and psychotherapy delivery."
Clinicians who treat mental illness are in urgent need of methods, tools, and data to efficiently track, assess, and respond to mental health needs throughout the treatment process. Similarly, patients need tools that provide feedback to optimize their therapeutic exercises.
"We propose to transform mental health assessment and care through enhancing these clinical practices with data-driven approaches."
I joined this project in August 2019, when I joined the Ubicomp Lab as a Graduate Research Assistant with Dr. Rosa Arriaga. My work focuses on designing and developing the patient-facing application. We are collaborating with Emory University for PE Therapy expertise and Rochester University for Natural Language Processing. This project is funded by the National Science Foundation (NSF).
Research
1. Understanding the Prolonged Exposure (PE) Therapy
The first step in understanding the PE Therapy was to read through the PE manuals used by therapists. The treatment and manuals are designed for use by a therapist familiar with cognitive behavior therapy and (CBT) or who underwent intensive workshops for prolonged exposure by experts in this therapy. The manual guides therapists and counselors to implement this brief CBT program that targets PTSD following various types of trauma.
Even though the therapy followed by each patient undergoing PE therapy is unique in some way, there is a lot of structure that has to be followed by the clinicians. After reading the manual, guide, and workbooks, I was able to formulate a standard PE workflow. A typical PE therapy lasts 9 days and here is a day in the life of a patient.
After understanding the PE workflow, I asked our primary contact at Emory University, a clinical psychologist who works with PTSD patients, if they used any existing technology for Prolonged Exposure Therapy - PE Coach. They told me that PE Coach (existing mobile phone application) had helpful features but was not used. So, I decided to investigate why.
2. The Outpatient Program at Emory
Even though I understood the basics of PE Therapy, there were some gaps in my knowledge. It was important to understand the specifics of the Outpatient Program at Emory. I conducted 3 semi-structured interviews with clinical psychologists who treat PTSD patients using PE therapy daily. I enquired about the following:
# Line of Questioning Purpose Summary of Answers
1 Feature Prioritization To determine the features necessary (high development cost) for PE therapy and add data collection. Necessary Features
  • Secure Login
  • Data Encryption
  • Manage Assignments
  • Record Sessions
  • Complete Homeworks
  • Visualize Data
Nice-To-Have Features
  • Add notes to In-Vivo sessions
  • Scrub through audio during Imaginal sessions
Distant Future Features
  • Breathing Tool
  • Reading List
2 Data Transparency To know about the different levels of data visibility for patients and clinicians. There should be complete data visibility for both patients and clinicians.
3 Homework Determination To understand who can assign homework to a patient At Emory, the patients control their therapy so both the patients and the clinicians can assign homework.
4 Time Dependency To understand how the hour of completion of an assignment affects the therapy The patient should have the flexibility to complete the assignment at any hour, but the system should know and display when it was completed.
5 Homework Flexibility To determine how if the application should impose various restrictions on the patient The system should impose minimum restrictions and provide as much control/flexibility to the patient as possible.
6 Audio Management To discuss how data should be stored and accessed on a mobile phone device The data has to be encrypted in storage and transfer. There will be no audio transfer because the cost of data infilteration/loss is too high.
3. Themes
After concluding my research, I used affinity modeling to combine the information and found 4 common themes.
SAFETY
The patients should feel safe while enrolled in the therapy.
TRANSPARENCY
The patients should be aware of the therapy process and should have complete access to all their data.
FLEXIBILITY
The patients can choose to complete assigned homework and can undertake previous or future tasks whenever they want.
CONTROL
The patients choose and assign their homework according to their comfort levels.
Design
1. System Architecture
I knew we needed to augment the PE Therapy using data so I tried to capture data from all facets of the patient's life without introducing any privacy concerns. The PECSS system encourages patients to share all data streams but they can choose to opt-out of as many data streams as they want. All the data captured is encrypted and not available to anyone except the patient and the clinician.
It is important to note that the data is captured only during the homework session. The architecture below describes how each data is collected and displayed. The patient is encouraged to use their smartphone and we provide them with a Fitbit to capture biological data. They also receive login details to a smartphone app to manage their therapy and the same login details can be used to view their progress on the dashboard.
2. Low-Fi
In terms of UI design, my job is to create the patient-facing interface for managing their therapy, i.e. the smartphone application. Based on all my previous research, I created the following Low-Fidelity Mockups.
I used to Mockups to conduct 4 remote think-aloud sessions with experienced psychologists across the United States. The tasks for these sessions were chosen to cover all the functionalities of the application. After the session, they were asked a series of questions concerning PE Therapy in their particular hospitals/centers. Here are the combined notes from 1 out of the 9 tasks that the clinicians were asked to perform:
Task Feedback Impact Category
Perform an Imaginal Homework Scrubbing is a good feature and should be included Improves flexibility Positive Feedback
Can scrub times be noted? Give clinicians more insight into patient's activity Think About
Add Session # in the title prevent users from going into the wrong session Immediate Change
Simple and easy flow Usability Positive
SUD values- need to be from 0-100, sliders and discrete options aren’t recommended, a simple number input would work best Ease of entering SUD values Need More Data
The scrubbing option is not clear Decrease Flexibility Immediate Change
The option to choose imaginal/complete is really nice Increase Control Positive Feedback
A traditional media player feel might be more helpful Increases Affordance Immediate Change
Fast Forward? Rewind? Skip by 10 secs? More features Think About
3. High-Fi
I received highly targetted feedback from clinicians on my first iteration, which has been crucial in helping me design the high fidelity version. This version will be tested with PTSD patients in Fall 2020. Here are some screens from the final design direction.
Development
I have been working on 3 core components in terms of development. While developing these features, it is important to note that all the data we collect should be secured on their device, during transfer, and the server.
# Component Purpose Language Platform/API Lines
1 Android Application The app is used by the patients to manage their therapy - Login, Add/View/Perform assigned homework, and upload data. It uses native phone sensors to collect user data, secure it using AES, and then upload it to the server using HTTPS. Java, SQL Android Studio 5,000+
2 SMS The patient provides phone numbers of 2-3 people close to them, we send them automated SMS texts to ask about their interaction with the patient. Python, SQL Twilio API 200+
3 Fitbit The Fitbit is used to collect biological markers of stress such as heartbeat, movement, etc. Python Fitbit API 100+
If you would like to know more about this project, reach out to me at [varnit@gatech.edu] or [LinkedIn] .