The development of Therapy Fidelity

Therapy Fidelity is a mobile interface that displays 132 sequential steps of cognitive behavioral therapy (CBT) for treating posttraumatic stress disorder (PTSD) to improve the fidelity of following the treatment protocol.

The Challenge

Leaders in the field of psychotherapy have been encouraging psychotherapists to adopt evidence-based treatments such as CBT for 20 years but it has been largely unsuccessful. Therapists have been unwilling and unable to rigorously follow CBT protocols with fidelity. Therapists cannot memorize all 132 steps, and patients have very limited access to the steps. As a result, treatment of patients suffers.

Child psychiatrist Michael Scheeringa, MD identified a solution for this problem—a mobile interface that displays a CBT treatment protocol and that can be viewed by therapists and patients simultaneously. Through the guidance of Dr. Scheeringa and in partnership with Tulane University, Agaric developed the Therapy Fidelity web app.

Our Approach

Designed For Use During a Therapy Session

When a patient attends their therapy session, it is far more likely that they arrive with a mobile phone than with a laptop. The need for a seamless mobile experience was one major reason we decided to build Therapy Fidelity as a single page application using React.js. Single page applications are great for reducing the number of network calls made to the host server since they do not need to request markup templates every time an end-user clicks to a different part of the site.

Consequently, most of the network calls made by the app are made to read or write data to one of several third party APIs. We worked to mitigate these, too, by using the React Query library, a powerful and configurable tool for fetching, caching, and updating data in React. The result is an app that clearly indicates when data is loading, allows the user to continue navigating the app while update requests are being processed, indicates the results of those requests once they resolve, and prevents redundant fetching of data that has already been cached—in other words, a fast, smooth, and intuitive end-user experience, even on a limited-bandwidth device. One additional benefit of the data caching—many third party APIs will bill per API call after a limited number of requests; without caching that data in the browser, these API services could get expensive fast.

Simple, Consistent, and Flexible Design Elements with Tailwind UI/CSS

Throughout the building process, we looked to Tailwind UI for example components. By doing so, we made sure that the app has a consistent look and feel and is responsive so that it looks good on devices of all screen sizes. As this app mostly functions as a utility dashboard, the pre-designed components sufficed for the apps aesthetic needs, meaning that Tailwind UI saved us a lot of time on design work. Often we had to make minor adjustments to the markup so that it would suit our specific needs, and very seldom did we need to add any custom classes to our CSS stylesheet, which incidentally results in removing the often expensive overhead of building and maintaining our CSS architecture. Again, this is great for an app with simple aesthetic needs but which also needs re-usable and flexible responsive components.

HIPAA Certified Storage with TrueVault

In order to facilitate a quickest time-to-market, we opted to utilize an existing HIPAA certified provider for the storage, retrieval, and access controls/monitoring of protected data. TrueVault provides a JavaScript library for communicating with their RESTful backend. Their backend includes provisions for dealing with all aspects of the data, such as authentication, authorization, logging, and so-on. We were able to use their Vault-based data model to separate patient records into individual silo's per-therapist, and use their group-based security to set nearly all relevant permissions for the app's use. TrueVault certainly comes with its limitations and challenges, but it sure would have taken quite a bit more time and resources to build and certify a custom storage system.

Automated Quality Assurance using Typescript and React Testing Library

Additionally, we needed to ensure that each client's data is stored and retrieved accurately. TypeScript, which provides a type system for JavaScript, helped to ensure that as our application passed data around, it was of the expected type. Using a typed language also reduces the possibility for software errors generally. Beyond using good tools to write the application, checks must be in place to ensure that as it evolves, things do not break. Here, we leveraged React Testing framework and other tools to walk through the functionality of the application and ensure that everything continued to function correctly, and even stored the correct data; By mocking (simulating) the TrueVault service we could inspect what the application would be storing there directly.

The Delicate Balance of Computing Large Sets of Sensitive Data

While the majority of permissions available to a logged in client were sufficient, there were cases where privileged tasks needed to be performed. Therapy Fidelity's People Like Me system is a great example. People Like Me involves a multi-step form which upon completion shows how many people with similar conditions were able to significantly improve their condition following the successful completion of a 3 month adherence to therapy protocol:

First, a patient selects for comparison the results of a single survey they have completed, consisting of numerical indicators  for several psychological conditions (anxiety, depression, PTSD, etc.)

Second, they configure which parameters they would like to match on. For example, a patient could choose to only see how many people who had shared a similar pre-therapy anxiety score were able to improve their condition following therapy.

Finally, they view the results.

Now it must be acknowledged that this system is parsing through and accessing patient data in real time. Of course, it would be a breach of patient privacy law for the patient to have such access to all of this other patient data, so we made sure that all data was scrubbed before reaching storage to reduce any personally identifiable information, and that computations occurred securely on a server before serving back results to the client.

Rather than utilize something like a full VPS server that performed these functions for us, we were able to leverage functions-as-a-service to provide the specific privileged function access, specifically AWS Lambda functions. This helped us keep the surface area of the application as low as possible for security purposes, while providing server-side style functionality. The language we chose for these was Go, for its performance, excellent built-in framework, and because we were able to build for our needs with very minimal additional libraries, which again helps mitigate security concerns.

While these functions-as-a-service were used several times throughout the site, it will be hard to forget Dr. Scheeringa's excitement about the realization of the People Like Me tool.

What's Next?

Therapy Fidelity was an excellent development experience—both for the thrill of developing a single page app and for carrying on Agaric's history of building tools for medical and scientific communities. As always, we are looking forward to our next opportunity to engage in these communities and build something wonderful.

For any clinicians who happen to be reading this and are interested in using the Therapy Fidelity app for their own practice, Scheeringa Mind offers it as a service and provides a 30-day free trial. There are also plenty of video tutorials on how to use the app, which might help to preview just what you might be getting into.

As Scheeringa Mind Company grows and Therapy Fidelity sees more use, we will be sure to publish any updates or major achievements. We are anticipating great strides in research on CBT protocol and look forward to seeing the results.