New Study Uses Brain Scans to Detect Hidden Heart Risk
                     Why a Hybrid Approach is Key for Bridging the Mental Health Data 
Gap
As health systems navigate rising costs, workforce strain, and an increasingly complex 
patient population, another crisis continues to deepen behind the scenes and it might 
surprise you. Despite an explosion of wellness platforms and increasing awareness around 
mental health, data from healthcare organizations reveal a troubling reality: many mental 
health disorders remain hidden in plain sight.
For health systems, this blind spot in mental health diagnosis carries significant clinical and 
operational risks. It’s not all bad news, though. With the right tools and processes, providers 
can finally access a more complete picture of the state of mental health. This can lead to 
improved care quality, outcomes, and safety. And it starts with the data.
Where Data Falls Short
Recent research led by Oracle Health, John Snow Labs, and the Children’s Hospital of 
Orange County  analyzed  over  109,000  patients  using  Oracle’s  Real-World  Data 
platform.  The  goal was to measure how well neuropsychiatric symptoms were captured 
using structured EHR data alone versus structured data combined with unstructured clinical 
notes.
The   gap   was   stark:   suicidality   and   self-harm   events   were   twice   as   common   
when unstructured data was included. Symptoms such as agitation, memory issues, and 
irritability
often appeared only in clinicians’  narrative  notes—missing  entirely from structured  fields. 
This underreporting is not just a data quality issue. It represents missed diagnoses, delayed 
interventions, and potential safety risks.
The High Cost of Structured-Only Approaches
Structured data—from diagnosis codes and prescription records, for exacmple—serves as a 
foundational  element  in  modern  health  systems.  However,  these  fields  often  exclude  
the nuances  of  behavioral  health.  Clinicians  may  intentionally  avoid  formal  diagnoses  
due  to stigma, uncertainty, or deferral to specialists. And what remains is formal codes 
without any clinical context.
By contrast, unstructured notes reflect the real-world dynamics of patient encounters. They 
can capture mood shifts, behavioral cues, cognitive changes, and patient-reported concerns. 
Without these narratives, care teams may miss early signs of certain conditions or overlook 
non-obvious symptoms that could shape treatment decisions.
This not only impacts patient outcomes, but can be a serious burden on hospitals and other 
providers. For healthcare leaders focused on quality metrics, readmissions, or value-based 
care, this information gap can lead to poor care planning and missed quality benchmarks.
NLP and LLMs are Forging the Path
Artificial intelligence (AI), specifically natural language processing (NLP) and large language 
models (LLMs) can enable health systems to bridge the data divide. By extracting insights 
from  unstructured  clinical  text,  paired  with  the  foundational  knowledge  we  have  of 
structured sources, the AI can surface hidden symptoms, detect trends across populations, 
and flag patients at risk earlier and more accurately.
According to the aforementioned study, the inclusion of unstructured data via NLP led to a 
20% increase in detected outcome events. This is a significant improvement that has a 
direct impact  on  clinical  decision  support,  care  coordination,  and  risk  stratification  
programs. Although the value is clear, implementation isn’t as cut and dry.
AI is powerful, but it’s not turnkey. To run properly, providers need annotated data, in some 
cases,   a   data   scientist,   clinician   validation,   and   robust   governance   frameworks   
to operationalize  AI  models.  Many  organizations  also  face  resource  limitations,  from  
staffing and budget to IT expertise.
Hybrid Models are the New Gold Standard
To meet the demands of real-world clinical environments, hybrid AI pipelines, which 
combine structured  rule-based  systems  with  flexible,  reasoning-capable  LLMs,  are  
emerging  as  the gold standard. These models not only extract clearly defined symptoms 
but also infer subtler signs of distress that would be missed by traditional methods.
Unlike  general-purpose  AI,  healthcare-specific  LLMs  can  adapt  to  clinical  language,  
learn from evolving documentation styles, and operate across diverse systems. For 
hospitals, this means scalable, high-fidelity extraction of behavioral health data from notes, 
regardless of EHR platform or documentation variance.
This  unlocks  a  powerful  capability:  automated,  comprehensive  mental  health  
surveillance that supports earlier diagnosis, improved drug safety research, and more 
precise population health management.
A Look Ahead
Millions  of  people  are  affected  by  mental  illness  each  year—and  these  are  just  the 
documented cases. In fact, nearly 60 million adults (23.08%) experienced a mental illness in 
the past year, according to research from Mental Health America. Among other worrisome 
findings,  nearly  13  million  adults  reported  serious  thoughts  of  suicide.  The  number  
of individuals who died by suicide in 2022 was the highest number ever recorded in the US.
The same research found that more than half (54.7%) of adults with a mental illness do not 
receive treatment, totaling over 28 million individuals. Even in states that ranked high in 
care, such as Montana (ranked #1), over 4 in 10 adults with a mental illness did not receive 
care. Again, this is only inclusive of the people who have reported experiencing symptoms.
To summarize, for now, mental health is getting worse, not better. In response, we need 
tools that  capture  the  full  clinical  narrative,  and  not  just  what’s  codified  in  structured 
 data. Integrating AI-powered analysis of unstructured notes into hospital workflows is no 
longer a forward-looking  experiment,  but  an  essential  piece  of  the  puzzle  for  
delivering  equitable, informed, and proactive care. After all, our mental health depends on 
it.
Author: David Talby
David Talby, PhD, MBA, is the CEO of Pacific AI, an AI governance company, and of John 
Snow Labs.  David  has  dedicated  his  career  to  making  AI  and  NLP  solve  real-world  
problems  in healthcare, life science, and related fields. He has extensive experience 
building and running web-scale software platforms  and teams  –  in  startups, for 
Microsoft’s Bing  in the US and Europe, and to scale Amazon’s financial systems in Seattle 
and the UK. David holds a Ph.D. in Computer   Science   and   Master’s   degrees   in   both   
Computer   Science   and   Business Administration. He was named USA CTO of the Year by 
the Global 100 Awards in 2022 and Game Changers Awards in 2023.
Company: John Snow Labs Inc.
John Snow Labs, the AI for healthcare company, provides state-of-the-art software, models, 
and data to help healthcare and life science organizations put AI to good use. Developer of 
Medical LLMs, Healthcare NLP, Spark NLP, the Generative AI Lab No-Code Platform, and the 
Medical Chatbot, John Snow Labs’  award-winning medical AI software powers the world’s 
leading  pharmaceuticals,  academic  medical  centers,  and  health  technology  companies.
Creator and host of The NLP Summit, the company is committed to further educating and 
advancing the global AI community.