A data breakthrough: Using CNLP to improve population health

Oct 06, 2022 Clinithink
A data breakthrough: Using CNLP to improve population health

Chris Tackaberry, CEO and founder of Clinithink, discusses how CNLP technology can assist healthcare organisations gain access to invaluable data insights that help improve population health and facilitate early intervention in a wide range of conditions.

Press article published on Health Tech World, 05 Oct, 2022

Non-communicable diseases including cancer, diabetes, strokes and dementia remain the cause of almost 70% of all deaths worldwide. Together they constitute another ‘epidemic’ which is putting healthcare budgets and frontline staff under enormous strain.

In the UK, where dementia is the leading cause of death, over 50% of cases currently go undiagnosed. This trend of patients living unknowingly with early-stage disease plays out across many non-communicable diseases – in the case of coronary heart disease, it is generally only discovered after an emergency hospital admission.

Improving detection of people whose chronic disease is about to get worse is essential for timely medical interventions, better clinical outcomes and patient experience, while alleviating the financial and time pressures on healthcare providers. In the UK a chronic disease such as diabetes can cost the NHS around £10billion pounds a year but with resources already overwhelmed, how can change in detection be achieved?

Leveraging data to create more effective healthcare

The first step is to embrace technology, creating systems that are cost-effective and sustainable. Specifically, technology that interprets an enormous amount of medical data, especially unstructured data, must form a cornerstone of the new healthcare ecosystem. This type of clinical analytics is invaluable for governments, healthcare organisations, researchers and anyone involved in the management and delivery of healthcare at scale.

With the adoption of electronic health records (EHRs), there has been an explosion of structured data. This information is typically captured in structured templates or forms, creating coded patient data that provides high level insights into patterns of disease and population health.

However, while some diagnoses are easily captured with coded data, there are severe limitations, particularly in relation to chronic disease. Here patients may exhibit a range of symptoms, most of which will not have corresponding codes. i.e.; patients with Minimal Cognitive Impairment (MCI) brought on by Alzheimer’s cannot be found with structured data as some of the most common indicators of the disease – from absent mindedness to irritability and muddled thinking – cannot be captured using structured codes.

However, EHRs also contain unstructured information held in the form of clinician notes, outpatient letters, discharge summaries and operation notes. This dense, unwieldy narrative comprises approximately 80% of meaningful data contained within the patient record. This is vital to the care process and is scrutinised frequently by frontline clinical staff. Yet, for analyses of any kind at scale, this information typically goes unanalysed, since most lack the technological ability to parse data in unstructured free text form. Accessing these insights through manual review can take months, years or even centuries-worth of precious clinical time.


The move towards AI-powered solutions

Healthcare technology powered by Artificial Intelligence (AI) can play a key role in solving this challenge. The vital clinical narrative contained in documents written by clinicians as an ongoing record of care can be unlocked with Natural Language Processing (NLP), a branch of AI that enables machines to understand the human language. Its goal is to build systems that can make sense of text and automatically perform tasks like translation, spell-check, or topic classification.

However, the language of healthcare is so complex that generalised NLP tools perform poorly in this space. Instead, Clinical Natural Language Processing (CNLP) incorporates an extensive clinical terms library, meaning the technology can understand billions of different word and phrase combinations that relate to hundreds of thousands of detailed clinical concepts.

By recognising and analysing the clinical and social data within this unstructured clinical narrative, CNLP-based technology can process lengthy, chronologically ordered content and make it computable at scale.

Leading CNLP solutions can read, “understand” and make millions of clinical documents computable in hours, identifying much higher numbers of possible patients that fall into patterns indicative of chronic disease.

Spotlight on: Diabetes

Diabetes UK predicts that there are hundreds of thousands of people living unknowingly with diabetes across the UK. Despite this huge burden of disease on health economies, it is largely controllable. Active management at specific stages in the disease process is high priority for healthcare providers across the world.

Diabetes is a complicated multi-system disorder creating a wide range of symptoms and clinical findings. Many of these characteristics are detailed concepts which cannot be coded, meaning that a reliance on structured data alone to find specific sub-groups of diabetics is challenging. CNLP can help to solve this problem by locating at-risk patients through an interpretation and analysis of long-form, unstructured data, giving clinicians the information they need to identify and assess vulnerable groups of patients.

It has been estimated that Diabetic Foot Disease (DFD) costs the NHS £2 billion per year in the UK. Barts Health NHS Trust, one of the largest trusts in the UK, turned to CNLP to help it locate patients at risk of amputation due to worsening DFD, with the goal of intervening in the disease process before amputation was necessary. Barts was able to identify over 6,000 patients with diabetes, of which over 3,000 had diabetic foot disease, in a matter of weeks - a process they predicted would have taken 100 clinical person years if completed manually.

Delivering Population Health

Demand for healthcare continues to increase, as do costs. The pandemic has exacerbated this trend by creating a huge backlog of non-Covid work. The number of physicians, nurses and other clinical professionals, cannot keep pace with the demand for services, which is why care-providing organisations must turn to new processes and technologies that improve efficiency, optimise quality care delivery and reduce costs. In the UK, the NHS is facing what has been called ‘the greatest workforce crisis in history’, with recent research showing that in England alone there is a deficit of 12,000 hospital doctors and over 50,000 nurses and midwives.

CNLP is a data breakthrough, affording healthcare organisations access to actionable insights previously hidden away, and ultimately helping to improve population health by facilitating early intervention in a huge range of important conditions.



About Clinithink

Clinithink’s platform delivers the deepest, fastest and most accurate analysis of unstructured data across patient populations currently possible. Whether it’s to better understand a vulnerable population, to highlight important phenotypes or to reduce the risk of denied claims for providers, Clinithink’s solutions provide healthcare decision-makers with better information and insight. Through automation, these insights can be gained in a fraction of the time of that manual processing would require and at reduced cost. For more information, contact: marketing@clinithink.com

Published by Clinithink October 6, 2022