Holmusk Announces New Publication: Natural Language Processing Models Based On Deep Learning That Translates Unstructured Notes From Psychiatry Into Quantifiable Actions

SINGAPORE AND NEW YORK, Jan. 28 / PRNewswire / The publication shows how Holmusk’s NLP models can convert unstructured psychiatric notes from EHRs into a structured, quantifiable format for analysis and comprehensive insight

Holmusk, a leading international data science and health technology company building the largest real-world evidence (RWE) platform for psychiatry, today announced the publication of the scientific article “Natural Language Processing-Based Quantification of the Mental State of Psychiatric Patients“, announced in Computational Psychiatry (MIT Press). Read the full text here: https://cpsyjournal.org/articles/10.1162/cpsy_a_00030/.

With this publication, Holmusk has validated its unique library of proprietary Natural Language Processing (NLP) models that translate unstructured psychiatric records into quantifiable indicators of patient status (e.g. symptoms, side effects, and external stressors). When these quantifiable indicators are used to enrich health systems data, they can assess patient disease severity across the spectrum of behavioral disorders and provide long-term histories of patient status. By creating these quantifiable indicators, Holmusk’s models generate, for the first time, solid real-world evidence of disease progression and treatment effectiveness for psychiatric disorders. Holmusk will use the objective metrics from these models to promote measurement-based care and individualize psychiatric nursing care across health systems.

“With the NLP labels we can generate structured information about the psychological state of patients from free-text notes from hospital doctors and quantify the severity of mental illnesses on the basis of these labels using deep learning algorithms. In addition, we can directly extract important psychiatric Labels from clinical notes circumvent the tedious process of anonymization, and this opens up interesting opportunities for us to apply quantitative analysis techniques to all available psychiatric notes, “says Sankha Mukherjee, Senior Data Scientist at Holmusk and lead author.

To create this library, Holmusk’s Data Science (AI) team used its proprietary electronic health record (EHR) system for mental health, MindLinc. The models convert more than 20 years of electronic medical records from Holmusk’s international MindLinc database into over 240 psychiatrically relevant dimensions. In combination with machine learning, these dimensions contribute to more precision in the daily management and treatment of diseases.

“Holmusk’s reliable and systematic compilation of physician records to evaluate outcomes across patients, physicians and healthcare systems is a huge step forward in a learning healthcare system. Based on patient care decisions and outcomes, we learn how best to care for other patients said A. John Rush, MD, Professor Emeritus at Duke-NUS Medical School, Associate Professor of Psychiatry and Behavioral Science at Duke Medical School, and Scientific Advisor to Holmusk.

Holmusk’s solid and scalable NLP models make it possible to extract information from unstructured EHR data from psychiatric treatments, a previously time-consuming and manual process.

“At Holmusk, we don’t wait to collect perfect data to analyze and generate insights. The ability to analyze the rich and varied information in doctors’ notes in a structured, quantifiable format is a potential turning point in the field For the first time, our approach allows us to segment patients not only according to diagnosis, but also according to clusters of symptoms and degrees of severity, which can then be matched to the appropriate medication in clinical trials and in practice, “says Joydeep Sarkar, Chief Analytics Officer at Holmusk.

Scientific sources: Mukherjee S.S., Yu J., Won Y., McClay M.J., Wang L, Rush A.J., Sarkar J. (2020). Natural Language Processing-Based Quantification of the Mental State of Psychiatric Patients. Computational Psychiatry, Volume 4, 76-106. https://doi.org/10.1162/cpsy_a_00030

Information on Holmusk:

Holmusk is committed to improving the lives of people suffering from behavioral and chronic diseases through evidence-based medicine. Headquartered in Singapore and with a global presence, the company drives innovation, research and care in healthcare and develops the world’s largest real-world evidence platform for mental and chronic illness. In 2020, Holmusk raised $ 21.5 million in its Series A funding round, led by Optum Ventures and Health Catalyst Capital.

Holmusk’s analytics platform synthesizes real-world data (RWD) using proprietary disease history models to generate actionable insights for research, innovation and care in behavioral medicine. Holmusk’s real-world evidence platform, NeuroBlu, is based on one of the largest longitudinal anonymized behavioral databases with data from more than 550,000 patients collected over more than 20 years and more than 20 million visits. Holmusk is continuously expanding its database through partnerships with health systems around the world. In addition, Holmusk develops digital health solutions that engage patients, support self-management of their illnesses, and capture patient-related outcomes to facilitate clinical decision-making and analysis. More information is available at www.holmusk.com

Inquiries & contact:

Paul Feldhausen
Director of Growth Operations


Read original article here

Leave a Comment