Talkspace Increases Access to Mental Healthcare Using Machine Learning on AWS

For many people, mental healthcare is virtually impossible to access. Approximately one in five Americans has a mental health condition, and over half of these individuals do not receive treatment for their conditions. Patients state that multiple factors—including lengthy wait times, high treatment costs, limited transportation options, social stigma, and lack of health insurance—prevent them from seeking care. To break down these barriers, digital therapy leader Talkspace created a solution: affordable access to professional therapists through web and mobile applications.

From its initial launch in 2011, Talkspace has used Amazon Web Services (AWS) to create innovative and secure behavioral health technology. Convenient and compliant AWS services provide many of the building blocks of Talkspace’s platform. “AWS assists us in delivering a HIPAA business associate agreement–level of service,” says Gil Margolin, chief technology officer at Talkspace. “And that’s a very important factor of AWS and one of the main reasons we chose AWS in the very beginning.”

Through the Talkspace platform, users can attend therapy sessions on their devices at the touch of a button—but matching patients with the exact care they need requires a certain level of sophistication. “Pairing each client with a mental health professional who meets their needs is important for the success of therapy. By standardizing our machine learning [ML] workloads on AWS, we’re able to understand our clients better, increase our level of service, and provide time-saving tools for therapists to improve client outcomes,” says Margolin.

Talkspace uses Amazon SageMaker and Amazon SageMaker Ground Truth to help improve the therapy it promises to provide. Amazon SageMaker, a fully managed service that makes it easy to build, train, and deploy ML models quickly, helps Talkspace match patients to therapists across multiple dimensions. Meanwhile, Amazon SageMaker Ground Truth, a fully managed data-labeling service, makes it easy to build highly accurate training datasets for ML. “Each time we improve additional dimensions, we create an AWS endpoint through Amazon SageMaker and are able to run A/B tests and improve our model,” Margolin says. “We refined this model with additional data points we ingest,and we’re able to improve the matching affinity score.” Amazon SageMaker has also helped Talkspace prevent early attrition by enabling Talkspace developers to create specialized models that continually improve the matching affinity score as clients engage with their providers. 

Outside the ML space, Talkspace continues to rely on tools like Amazon Elastic Compute Cloud (Amazon EC2) for resizable, secure compute capacity and Amazon Elastic Container Service (Amazon ECS) to run sensitive and mission-critical applications, like its matching algorithm. AWS Lambda enables Talkspace to run code without provisioning or managing servers, automatically facilitating fast, consistent data processing. By continually capturing important data to predict user engagement and by measuring the data against patient outcomes, Talkspace was able to refine and optimize its models to improve the effectiveness of its therapy.

Talkspace uses ML to identify behavioral patterns and potential harm risks, sending push notifications to therapists in real time if elevated risk is detected. The company uses Amazon SageMaker and Amazon ECS to develop a diagnostic profile of each patient and provide helpful insights, such as potential secondary conditions, that therapists can review. “We received very positive feedback and reports of satisfaction on this model from the majority of therapists,” Margolin adds. 

Additional ML-assisted features provide therapists with recommended tips and actions to avoid early dropout and to retain patients, as well as space to reflect and adjust treatment courses. Talkspace also implemented autoprogress notes to enable therapists to quickly create session reports and summaries and easily visualize a patient’s progress over a specific period. Due to the time-saving benefits, the company saw a 20 percent increase in the number of notes created by therapists after implementing these features.

The Talkspace team also relies on Amazon Transcribe, a tool that makes it easy for developers to add speech to text capabilities to their applications. “We could have gone and transcribed audio messages into text, but this is something that AWS helped us solve, and again, that saved us a lot of time,” Margolin says.

Talkspace has big plans for the future. Having reached over two million people to date, it hopes to integrate ML vertically across the continuum of care, providing every tool under one roof for therapists and patients alike. Talkspace expects to use new AWS services to improve its existing algorithms and continue to evolve, providing more patients with holistic, collaborative, and integrated mental health treatment. “AWS is able to assist us in focusing on the things that really count: our core mission, our core product, and our core competency—delivering quality therapy at scale to people who deserve but are unable to receive that level of care,” Margolin says. “This was made possible with the services we created on the foundations of AWS.”