Dr Chris Tackaberry of Clinithink explores how making sense of unstructured clinical information can help the NHS make the most of what it has already has
Making the most of existing assets is a must for healthcare leaders, and technology and data are some of the most-valuable ones they possess. However, there is rich intelligence held in these assets that can often be overlooked. Dr Chris Tackaberry, chief executive of Clinithink, explores how making sense of the large amount of unstructured clinical information can help the NHS make the most of what it has already got
At a time of increasing demand and dwindling resources, organisations such as the NHS have an opportunity to exploit two of the most -aluable assets they already hold, that is information about those who use its services, and the technology to record this.
NHS hospitals are able to use an electronic patient record (EPR) as the system to bring these assets together. Such technology gives clinicians easy access to vital patient information at the point of care. Doctors want to know everything relevant to make the right decision, and that information needs to be available quickly.
At a time of increasing demand and dwindling resources, organisations such as the NHS have an opportunity to exploit two of the most -aluable assets they already hold, that is information about those who use its services, and the technology to record this
But are doctors seeing everything they need to know? Patient information comes in many forms – patient notes, referral letters, relevant reports, patient wishes - a mixture of unstructured narrative, and coded data that, while electronically accessible, can be difficult to bring together in a timely, coherent fashion.
For healthcare, this is a significant issue. Industry experts estimate that 80% of clinically-relevant information may be held in such unstructured form. If this data cannot be used when needed, diagnoses can be missed. The most-effective treatment plans may not be put in place and valuable insights can be left at the bedside.
For example, patients with congestive heart failure will often have had echocardiography, which, among other things, will provide an estimate of Left Ventricular Ejection Fraction (LVEF) – a useful index of heart failure disease status and progression. Typically, this will be entered as free text in the electronic echo report and potentially cited in the patient’s case notes to support decisions made regarding treatment. The clinician can act on this information when dealing with an individual patient. However, this information could be used to the benefit of a wider health population. The challenge here is that unless the LVEF is recorded in structured form, a wider analysis of any kind is a matter of chance and not wholly reliable. Yet we know today that it is usually recorded in an electronically-unstructured rather than structured form in the record.
Converting rich narrative but unstructured data into something structured that can be utilised is therefore important. For instance, in the LVEF example, we can more easily identify those patients who have a certain ejection fraction, but who are not on the optimal medication shown to improve outcomes and reduce the likelihood of progression to more-serious disease status.
A deeper, technology-enabled understanding and structuring of natural clinical language expressed by clinicians in narrative form can enable more informed clinical decisions, improved clinical outcomes and reduced cost
Pro-actively identifying the cohort of patients that meet these criteria enables early intervention in the management of a disease for not one, but multiple patients, reducing the cost of caring for the group as a whole and improving outcome and quality of life to boot.
Strategies of this kind are no longer optional for the NHS; they are mandatory, offering both reduced cost and improved outcome in a single approach. They deliver benefits for both the individual patient looking to receive the best-possible care for their personal needs and the healthcare provider looking to make the most of hard-pressed healthcare resources.
Doing so in this and many other examples depends on being able to access and analyse unstructured clinical data where the information needed to drive actionable insight actually sits.
Numerous solutions are available to support analytics driven by structured data captured in EPRs and related systems. These are important components of a strategy that seeks to make the most of information assets in redesigning care services, targeting healthcare resource where it can be most effective. However, as we have seen in the example above, it is important to leverage unstructured data as well, providing a more-complete picture and recognising that in many cases, the granularity of clinical information needed to provide the necessary insight is only available in narrative.
Clinical Natural Language Processing (CNLP) makes this happen. It uses technology to make sense of unstructured narratives held within EPRs and elsewhere, and puts the narratives in a more-structured and understandable format.
What emerges is something extremely useful and powerful, as IBM’s Watson computer is showing. The artificially-intelligent computer system is using CNLP with Memorial Sloan Kettering Cancer Centre in the US to see if it can analyse anonymised cancer patient notes and records and turn them into actionable clinical practice to improve resource utilisation.
At Clinithink, we are using CNLP techniques and related products with Mount Sinai Hospital to help provide real-time guidance to clinicians on prescribing strategies tailored to a patient’s genome.
Both are examples of how a deeper, technology-enabled understanding and structuring of natural clinical language expressed by clinicians in narrative form can enable more informed clinical decisions, improved clinical outcomes and reduced cost.
These examples represent revolutionary, patient-centred healthcare capable of being delivered at scale, but with direct application in clinical workflow. CNLP is taking existing intelligence and existing health technology and transforming it into something that can help deliver better at lower cost, – the goal for healthcare systems across the world.
Clinical Natural Language Processing is taking existing intelligence and existing health technology and transforming it into something that can help deliver better at lower cost, – the goal for healthcare systems across the world
In this context, the argument for leveraging the powerful, untapped asset of unstructured data is strong. I argue that it is an asset that can no longer be ignored. It also has relevance when it comes to healthcare planning.
Clinical commissioners and public health planners will recognise that risk stratification is an increasingly-essential part of the planning process. As the Kaiser Permanente pyramid shows us, there are different levels of healthcare need within particular groups. Small cohorts, such as those with uncontrolled or worsening chronic disease, can use up large amounts of healthcare resources, especially when they are not explicitly known to the healthcare system from a holistic perspective.
Understanding those groups is essential for planning and resource allocation. That understanding can only come from as complete a picture as possible. This information can enable clinical commissioners to plan and target the services that most effectively, and efficiently, meet the needs of their communities.
Work is underway to look at how CNLP can support this planning process. Just as clinicians can benefit from seeing the whole picture about an individual patient, commissioners can benefit from seeing the needs of whole groups. Naturally privacy and consent are central to this work.
But this should not be confused with big data. This is rich smart data – comprehensive and comprehensible information that can directly inform patient care and service planning.
The benefits are clear; a personalised approach to care that taps into health information already recorded. Healthcare planning can be based on a more-accurate representation of the needs of groups of patients defined by complex granular clinical criteria. /
The digital revolution for patient information has come. As the use of EPRs and related clinical documentation systems grows and grows, so will the amount of data they hold, structured and unstructured. The challenge is to make sure all of that data is useful and useable
This has the potential to transform healthcare, both for the individual and for society at large. This level of innovation may seem daunting. Traditional ways of accessing patient information, such as paper notes, have worked for many years. True, they are excellent for one-on-one patient care, but they can only hold so much information, and only a fraction of the information available is there.
EPRs represent a step forward in bringing together important and diverse kinds of information. The digital revolution for patient information has come. As the use of EPRs and related clinical documentation systems grows and grows, so will the amount of data they hold, structured and unstructured. The challenge is to make sure all of that data is useful and useable.
CNLP is one area rising to that challenge. It is enabling healthcare systems to make the most of an asset that many have in abundance – data. And, by making sense of the rich information recorded in clinical narrative, and presenting them where they can have the most impact, health systems have the opportunity to improve outcomes and reduce costs.
It is an area where innovation is generating clinical and financial benefits and it should be high on the agenda for healthcare leaders looking to do more with less.