By Valentin Tablan
Over the past decade, technology has transformed the way we live. The emergence and advancement of personal devices, data capture, networks and the cloud, deep learning and Internet of Things (IoT)—amongst other things—has changed our world. Its impact is everywhere, including in the way we deliver treatments to patients through digital healthcare and therapeutics.
By combining this technology with the data we have at our disposal, we can accelerate the way healthcare operates and works to treat more patients globally. For example, by applying artificial intelligence and deep learning techniques to data, we can unlock its potential by using algorithms to see patterns in clinical outcomes and patients’ behavior. These can be used to inform improvements to today’s interventions and build new digital treatments that can be accessed anytime and anywhere, and made available to people or parts of the world who don’t have access to healthcare today.
Already today, we are seeing technology being used to enable a wide range of digital healthcare apps and services that are marketed as the new digital cure, particularly in the wellness space. However, even though a majority of these are apps are designed to treat milder conditions, such as anxiety or stress, they don’t always treat the patient’s needs as effectively as human or medicinal intervention, and are by no means a standalone treatment.
Of the digital treatments that are out there, only some are supported by scientific evidence such as clinical trials, and even those in some cases fail to deliver as effectively in the real world due to a lack of representation in the trial population, or mismatches between tightly controlled trial conditions and real life.
While clinical trials establish safety and efficacy, to get digital treatments to a place where they really deliver true value to patients, treatments should really be reviewed and assessed on real-world populations, outside the context of a traditional tightly-controlled clinical trial.
By taking large amounts of consensual patient data, and cross-referencing it with other datasets, as an industry, we can start to develop a clear picture of what works and what doesn’t in treatments and therapy. With this insight at our fingertips, we can then enhance current medical interventions and build new digital healthcare services that get more people better faster. Advanced data science techniques, including deep learning, allow us now to continually learn from data and apply that to the development of digital therapeutics.
Data that determines digital outcomes
In some settings of mental healthcare, this real-world data is already being gathered from the outcomes and interactions between patients and practitioners. With the participation of patients, clinicians, and scientists, we are able to apply data science techniques to data gathered over many years in order to study patient characteristics, interventions, and outcomes. These insights help clinicians to direct people in need to the best possible treatments, while also building a new generation of digital treatments that will ultimately treat patients as effectively as traditional medicine.
Ieso Digital Health recently published a paper (JAMA Psychiatry) presenting its “first of its kind” research using a deep learning model to quantify the impact of the therapists’ language during psychotherapy sessions. This work was done by analyzing over 90,000 hours of internet-enabled Cognitive Behavioral Therapy (ie-CBT) transcripts, and correlating that with clinical outcomes for the patients. The scale of this work means it would have been impossible for humans to do alone, and it is one of the examples of how modern technology opens new avenues of scientific inquiry that were not accessible before.
The study demonstrated the potential of what could be achieved if this type of data and research method is applied across the board, providing a first step towards practical mental healthcare that is actively quality-controlled. The same results give a strong indication of what may be possible, taking steps toward digital therapy driven by algorithms based on real-world evidence.
Decoding therapy to drive global digital healthcare developments
By analyzing patterns of symptoms in tens of thousands of patients we are discovering what impacts people and where there are common experiences. We can examine how each word and every idea expressed helps or hinders recovery from illness. We are able to associate each therapist and patient “utterance” with high-level clinical concepts and really learn which words need to be said when for each person.
Decoding therapy in this data-driven way, and using that to better understand and improve human-delivered care is just the start of this revolution. For areas of the world where there are few therapists, we may be able to ultimately use these datasets to train computer systems to perform some elements of personalized therapy with minimal or no human intervention. These technologies and statistical models developed from real-world evidence can be used to support people in need wherever they are.
This progress is invaluable to the healthcare systems of the world, who are faced with the compound pressure of increased demand, lack of investment, scarcity of therapists and limited resources, and will increasingly need to rely on effective digital treatments to provide high-quality mental healthcare to all who need it.
For the patient, this could also mean access to anytime, anywhere virtual therapy all from an app on a mobile device, through a conversational assistant, or even from a wearable device. An individual suffering from anxiety, for example, could trigger their virtual therapist by voice, and get the help and treatment they need when they need it. All powered by machine trained digital therapeutic services, this virtual therapist could emulate the very best in human care delivery, providing personalized 1:1 therapy and intervention to the patient on demand, or when the app detects a patient’s change in mood.
The opportunity bought by algorithmic diagnostics and therapy for widespread adoption is exciting and can address health conditions globally—all by extracting current knowledge from today’s standard clinical practice. Machines can learn to develop effective treatments for many mental health disorders, meaning patients get direct support, and removing the barriers associated with treatment today. It also means the precious few therapists we have available today can focus on the patients that need them the most.
We are not just on the crest of a new wave of digital therapeutics, we are witnessing the birth of a new way to do medicine. This will change the world!
About the author:
Dr. Valentin Tablan has spent 20 years in the field of Natural Language Processing, Knowledge Representation, and Artificial Intelligence. As SVP AI, at Ieso, Valentin and his team are responsible for applying advanced deep learning techniques to Ieso’s vast data set to determine which evidence-based therapy interventions are most effective. Under his leadership, Ieso has created the industry’s first AI-enabled tools that augment real-world data to increase quality and improve clinical outcomes.