Team members from Holmusk recently had the opportunity to attend the virtual annual Going Digital: Behavioral Health Tech conference and share more about the capabilities of NeuroBlu, Holmusk’s flagship data and analytics platform, along with insights gleaned from recent studies that have leveraged NeuroBlu.

Holmusk’s Chief Medical Officer Scott Kollins, PhD, spoke with Rashmi Patel, MD, PhD, Holmusk’s VP for Medical and Scientific Affairs about recent progress and future plans for the company as it continues to deliver on its mission of transforming behavioral health through real-world evidence. Highlights from their talk included:

‍

NeuroBlu in the Literature

Patel recently led the development of a peer-reviewed manuscript that detailed the data available within NeuroBlu, a real-world data and analytics platform that enables the generation of real-world evidence. One of the key indicators of the company’s rapid growth is that the platform included information on about 550,000 patients when it was submitted–and less than a year later, that number has doubled to nearly 1 million patients.

NeuroBlu contains information such as demographics, diagnoses, medications being prescribed, and–importantly–outcomes data.

“The real strength of our data is the availability of outcomes data recorded in real-world clinical practice,” said Patel, adding that this allows researchers to conduct longitudinal studies that take into account how well patients are doing.

‍

Natural-Language Processing Opens Doors

Kollins and Patel also discussed the proprietary natural language processing (NLP) models that have been developed by Holmusk and leveraged within NeuroBlu.

While NeuroBlu contains many structured data points, such as specific outcomes assessments, it also contains a vast amount of unstructured data, or information produced when a clinician types a note that could include details such as patients’ functioning and presentation, their symptoms, and their treatment response.

The NLP models are used to extract clinically meaningful information, such as symptoms or social factors, from these unstructured data.

“It is this kind of unstructured data that is a gold mine and it holds very rich clinical information,” Patel said. “The ability to extract and assemble that information automatically with natural language processing really transforms the availability of the data to allow us to generate at scale evidence that would simply not be possible otherwise.”

‍

Common Data Model Enables Generalizable Research Results

The conversation provided an overview of Holmusk’s work to harmonize data from many disparate sources, translating it into a common data model to enable comparison of data from many different clinics and centers.

It is this data harmonization, Patel said, that enables studies to include enough data points to ensure their results are generalizable beyond one center or one group of patients.

Kollins likened data to raw minerals and the data harmonization process to mining. “Data are not that valuable without the infrastructure to extract and process them, and that’s where Holmusk is unique, specific to behavioral health data,” he said.

‍

Data Privacy is Paramount

Patel also spent time discussing the de-identification process that data go through before they reach the secure, web-based NeuroBlu platform.

Data are de-identified at the source using the Safe Harbor method, meaning that identifiable information never enters into NeuroBlu. Furthermore, users of NeuroBlu are not able to download data from the platform, and all analysis happens within the secure platform.

“It’s tremendously important to safeguard confidentiality and the security of the data, and that’s something we take very seriously,” Patel said.

‍

Recent Behavioral Health Insights

Patel and Kollins also shared findings from recent studies conducted using the NeuroBlu platform. These studies, which have also been shared over the last month at academic conferences, include: 

• A study that examined outcomes in individuals who have schizophrenia found that those patients who have comorbid substance use disorders were more likely to experience worse outcomes across the board. This analysis of outcomes was feasible because the NeuroBlu platform contains an important outcome measure, the Clinical Global Impression Severity (CGI-S) scale, for most patients. Patel said these findings have important implications for health care policy, as solutions that address both schizophrenia and substance use disorders are more likely to be effective in improving outcomes.
• A study that used NLP to extract detailed clinical information on symptoms experienced by patients with schizophrenia found that patients with heavier symptom burden were more likely to experience worse outcomes. This will be important, Patel said, for both the development and delivery of treatments and helps to enable precision psychiatry, as patients with more symptoms will need targeted treatments to determine what works best for them.
• A study that analyzed how patients with ADHD functioned in relation to how many medications they were prescribed found that improvement in functioning was not associated with the number of medications a patient was prescribed. Kollins, a clinician specializing in ADHD, noted that insights like this are critical when treating patients and trying to achieve outcomes that matter most to them, such as improved functioning.

‍

Next Steps for Holmusk

Kollins and Patel also spent time discussing next steps for the company. In the near future, Holmusk plans to:

• Continue growing its data asset in order to generate even more insights that will impact behavioral health research and care delivery.
• Develop clinical analytical tools that will be designed to inform clinicians’ decisions with data.
• Explore opportunities to engage directly with patients in clinical research.

‍

‍