Modernizing Models of Care for Psychology (2024)

Using online, self-guided, low-intensity support to make a difference
Diya Shah & Shreya Ragade

Awards: Beaverton Science EXPO 3rd Place 2024(Computer Science Division)
Northwest Science EXPO 2nd Place 2024 (Behavioral Health Division)

2. ABSTRACT:

As the demand for mental health care increases in the United States, many patients find themselves in limbo as they wait for their needs to be assessed. Hence, an intervention in the form of an asynchronous online program could be a valuable means of mitigating the demand. Our proposal works to address the concerns of traditional therapy methods by creating an almost completely digital implementation of evidence-based practices, minimizing the need for human intervention. Upon the conclusion of the program, a treatment plan can be determined, whether that be a 45-minute therapy session known as an SSI meant to provide immediate care, or successful completion of the program, obviating the need for traditional therapy. 

We have developed a framework which performs sentiment analysis of comments in a forum, allowing a clinician to compare it to manual survey results.  The framework is based on free/open-source software and can be run locally on a provider’s computing infrastructure, which makes it straightforward to guarantee patient confidentiality. This gives clinicians a holistic view of the patient’s progress without having to individually meet with each one. Overall, alleviating the stress clinicians face when they are overloaded with patients is our primary goal, along with deepening our understanding of Sentiment Analysis and Data Cleaning. The final solution, a scalable implementation of evidence-psychological practices combined with a data visualizer poses a breakthrough in the intersection between psychology and computer science. 

3. IDEA

Problem: 

In the aftermath of the COVID-19 Pandemic, mental health cases in the United States have surged, leading to overwhelmed therapists and extensive waiting lists for those seeking care. Long wait times for mental health care pose significant challenges for families seeking support. Delayed access to therapy can worsen existing mental health conditions, leading to increased distress and heightened risk for more symptoms. Prolonged waiting periods can intensify feelings of helplessness for parents trying to improve their relationships with their children. Additionally, prolonged wait times might lead to a decreased likelihood of seeking help. Addressing and minimizing these wait times is critical to ensuring timely and effective mental health support, fostering better mental well-being, and reducing the potential for escalating mental health crises.

Current Work:

Modern solutions primarily include the SSI (Single-Session Intervention) which revolves around a single, 45-minute therapy session intended to “enhance the efficiency and cost-effectiveness of evidence-based practices in healthcare” (Jessica Schleider, 2022). Namely, they intend to address the immediate needs of patients who don’t require long-form therapy but could also benefit from medical intervention. However, SSIs are structurally designed for those who urgently need assistance, which all patients don’t require, and could pose a waste to those who would benefit from low-intensity support over some time rather than a single intervention.   

Solution:

Known as the Be Our Guest Program, our proposal works to address both the concerns of traditional therapy methods and the SSI by creating an almost completely digital implementation of evidence-based practices, minimizing the need for human intervention. Every week sees new modules with information relevant to patient struggles — in our case parenting. Participants can then utilize our forum to share their experiences and thoughts about the content provided. The comments left by participants are then collected and analyzed via a sentiment analysis software. Next, a survey relevant to parenting struggles is sent out weekly and scored accordingly. The score collected from the survey is then compared to the score received from the sentiment analysis of the comments, before being displayed in a dashboard-style format (Power BI) which emulates a typical clinician’s -heads-up display. This allows clinicians to analyze and give patients both subjective and objective input as to how they are improving. It also eliminates the cost-heavy need for direct patient-therapist interaction and can be scaled to address the needs of as many patients as possible over as many weeks as there are modules present. The program concludes with either successful completion or an SSI, depending on the patient’s progress. 

Image 2: Current sequence of the Be Our Guest program

4. PLAN

Resources: 

As far as physical resources go, only laptops are needed because our project is purely computational.

Goals:

Our goals can be broken down into three stages. The first part is the creation of a database with the patients’ comments and their associated sentiment value. Our second stage involves finding normalization values for our survey. The final part is the graphical representation of both of these values in a heads-up display. Below are our milestones:   

1. Creating a database with the Lemmy comments and Sentiment Analysis scores. Lemmy is an open-source comment forum which can be self-hosted. The task of the Lemmy instance is to provide a framework from which patients can talk about their responses to the week’s prompts and get feedback from clinicians and fellow patients. Clinicians must receive these scores in a neat format so that they can quickly pinpoint patients who may be sharing negative thoughts and ideas.      
2. Displaying the data from the CSV file and the survey data on Power BI. The end goal of this project is to have a complete graphical interface that showcases the patients who may require more care than others. To do this, we need to continuously update the Power BI software with new sentiment analysis scores, and new survey data results.  
3. Receiving feedback from mentors about technical/computational improvements. Also, testing our final creation on real patients from Oregon State University (OSU) and getting their feedback about whether they would be inclined to try this form of therapy.  

Our project will be tested by real patients from a psychology clinic at OSU. They will complete a survey every week while also adding comments on our Lemmy instance. Over the course of the program, we will display the collected data in Power BI, and show the clinicians how the results will look. From there we will collect their feedback, and make any improvements as suggested. Our project’s performance specifications are that it must be used by a parent who is looking to improve their relationship with their children.

Partner Collaboration 

Both of us have worked together during the Summer (Institute for Computing in Research), and have acquired many communication skills. One of us will focus on the graphical interface and the survey data, while the other one will work on the sentiment analysis model and the database. In addition, we will have a shared Google Drive and GitHub account where we will upload reports and code for each other.  

Risks:

Possible points of concern:

1. VADER sentiment analysis uses a dictionary of words, each with a sentiment value attached to it. A common drawback of VADER is that it uses dictionaries to assign a negative or positive connotation to a specific word. Therefore, this model is unable to understand if a negative word was used positively. To ensure the appropriate sentiment value is assigned, we will try creating our model using a dataset that we create, as opposed to a dictionary.  
2. The asynchronous and digital nature of the program, eliminating all human-to-human interaction may cause patients to be more reserved with their comments and feedback on the forum, which in turn could lead to less accurate understanding of their current mental state. Although this is inevitable, the nature of the program allows for more patients to get treatment, which is better than not receiving anything at all. A SSI is always a viable option for patients if necessary.
3. One of the main aspects of fully implementing this project is having Human participants test out both aspects of our project. However, we must ensure that we get explicit consent from each of our participants and the organization we are working with. To ensure that we do not face any consent issues, we will fill out an IRB form and any other forms required for human participants. 

Approach & Timeline: 

Our project is feasible because it improves upon technology that has already been developed. It incorporates these different technologies and allows them to communicate with each other to produce one cohesive result. 

Steps to implement our project proposal/ Sample Timeline:

1. 2024-02-14: Finish creating a program to transfer the comments and sentiment analysis from the terminal to a CSV file which is then automatically updated every day
2.  2024-02-29: Experiment with creating our own sentiment analysis model using BERT classification or another model
3. 2024-03-31: Normalize the data from the surveys and figure out a way to compare those to the sentiment analysis scores
4. 2024-05-12: Use PowerBI to visualize all the data in different formats (Highlight patients who may require an SSI, Provide individual patients progress charts, etc) 
5. 2024-05-31: Switch to a unique cloud server and use this project in the real world
6. 2024-06-01: Finalize GitHub account with relevant code.   

GitHub is our primary means of documenting our progress with which we will upload versions of our code as we make them. We will also type up a summary of what we hope to accomplish every day, what we accomplished, and any code associated with that.

Current Progress:

Earlier we mentioned a Lemmy instance we have already created. We have software ready to collect comments from its database and apply sentiment analysis on them.  We also have prepared algorithms based on trial posts to the forum. Below are images of the current system:

Image 4: Current Lemmy Instance

Image 5: Current organization of comments from the forum & Sentiment Analysis scores

Project Budget:

Figure 1: Cost Analysis of the Resources Required for our Project

Item	Amount to be purchased	Cost	Link to suppliers	Comments
Power BI	Power BI Premium	$20 per month	https://powerbi.microsoft.com/en-us/pricing/	Used for displaying sentiment analysis & survey data.
Google Domain	1	$20	Lemmy.computinginresearch.org (current)parentsupportwall.com (future)	To host our Lemmy instance.
Testing Project	N/A	$300 – $500	N/A	Funding required to test our project on real patients.
GPU Processing Bits	Colab Pro +	$49.99 per month	colab.research.google.com	To create a new Sentiment Analysis model.

References: 

Ahuvia, I., Mullarkey, M. C., Sung, J. Y., Fox, K. R., & Schleider, J. L. (in press). Evaluating a
Treatment Selection Approach for Online Single-Session Interventions for
Adolescent Depression. Journal of Child Psychology and Psychiatry.

NumFOCUS, Inc. (n.d.). User Guide. Pandas documentation. Retrieved December 25, 2023,
from https://pandas.pydata.org/docs/user_guide/index.html 

Maity, A., Wang, A., Dreier, M., Wallace, V., Orchard, F., Schleider, J. L., Loades, M., &
Hamilton, J. L. (in press).  How do adolescents experience a newly developed Online
Single Session Sleep Intervention for adolescents?: A Think-Aloud Study. Clinical
Child Psychology and Psychiatry

Schleider, J. (2021). Lab Projects. Lab for Scalable Mental Health. https://www.schleiderlab.org/currentprojects.html

Smith, A. C., Ahuvia, I., Ito, S., & Schleider, J. L. (in press). Project Body Neutrality: Piloting
a digital single-session intervention for adolescent body image and depression.
International Journal of Eating Disorders

Sung, J. Y., Buggati, M., Vivian, D., & Schleider, J. L. (in press). Evaluating a telehealth
single-session consultation service for clients on psychotherapy wait-lists. Practice
Innovations. 

The PostgreSQL Global Development Group. (n.d.). PostgreSQL 16.1 Documentation.
PostgreSQL: The World’s Most Advanced Open Source Relational Database. Retrieved
December 27, 2023, from https://www.postgresql.org/docs/current/

Wolraich, M. L., Hagan, J. F., Allan, C., Chan, E., Davison, D., Earls, M., Evans, S. W., Flinn, S. K., Froehlich, T., Frost, J., Holbrook, J. R., Lehmann, C. U., Lessin, H. R., Okechukwu, K., Pierce, K. L., Winner, J. D., & Zurhellen, W. (2019). Clinical Practice Guideline for the Diagnosis, Evaluation, and Treatment of Attention-Deficit/Hyperactivity Disorder in Children and Adolescents. American Academy of Pediatrics. https://publications.aap.org/pediatrics/article/144/4/e20192528/81590/Clinical-Practice-Guideline-for-the-Diagnosis