Life without smartphones seems nearly unthinkable today, and the reason why is simple: They facilitate countless tasks that ease daily life. The question being asked by many healthcare providers — and rightly so — is how to achieve similar benefits by maximizing the power of EMRs to continually drive improvements in patient care delivery, outcomes and efficiency.
Although a recent study found that EMRs do have a positive impact on outcomes, too often providers are limited by the ability to use only the information that is entered into the EMR as structured data, but not the free text.
Consider, for example, we know that patients with congestive heart failure account for a very high-risk, high-cost and vulnerable patient population. Yet much of the time, clues that identify them as such are recorded within the clinical narrative as free text —within an echocardiogram report or a clinician's note indicating reduced ejection fraction, for instance. As these patients travel across the continuum of care, many providers are unable to access these golden nuggets of information. As a result, a CHF patient might be at greater risk of ending up in the emergency room or being readmitted. Proactively identifying these vulnerable groups for targeted intervention is likely to improve outcomes, while at the same time reduce cost, according to an Agency for Healthcare Research and Quality study.
The clinical narrative routinely documented as part of a patient encounter often is described as an EMR "black hole." Critical information is recorded in the EMR, but not as standardized data that is searchable and quantifiable. Therefore, it is difficult to analyze and query this data, and get it back out of the EMR to be used for clinical decision support, analytics and other critical activities that enable better individual patient care and population health management.
As healthcare organizations move toward population health management and value-based care, they need to gain more insight into patient information and metrics for improving care delivery. An evolving technology known as clinical natural language processing can help organizations analyze and mine unstructured data from within the EMR's clinical narrative to derive the most value from the capture of both historical and current patient data.
Why an EMR isn't enough
Over the last few years, with the HITECH Act and Patient Protection and Affordable Care Act providing incentives to implement EMRs and penalties for failing to do so, healthcare organizations have invested millions in EMR technology. A lot of positive changes have resulted from these investments. EMRs have sped up data-collection efforts, added momentum to plans for engaging patients in their own healthcare and spurred the growth of initiatives such as e-prescribing. EMRs have also helped physicians gather large amounts of structured data that can be securely stored, organized, aggregated, analyzed and shared.
Yet EMR systems are hamstrung in their ability to do anything of consequence with the clinically rich unstructured data situated within their black holes. One prime example is the information gathered about patient allergies. Because this type of information is highly personalized, it can be complicated to capture using templates and difficult to store in a structured format. For each individual patient, it is important to document the general allergy, the specific nature of the allergic reaction, the precipitating agent, any mitigating factors, etc. As this vital information is collected across the continuum of care within different EMR systems, the details stored only as free text become trapped within separate electronic silos. This poses the potential for grave consequences for each individual patient, as well as negative implications for the broader patient population.
In today's healthcare environment, organizations must have access to both structured and unstructured data if they wish to elevate the level of patient outcomes in the most efficient and cost-effective manner. Reducing readmissions and encouraging preventive care, for instance, can positively impact an organization's bottom line and the quality of patient care. However, initiatives such as these are only possible when organizations can fully mine their data to identify at-risk patients.
How clinical natural language processing offers value
Despite the limitations of the EMR with respect to using unstructured data, it is possible to include CNLP technology that utilizes clinical narrative to aid providers in real-time at the point of care and positively impact patient care. A highly scalable solution, CNLP works by using an advanced clinical language-indexing engine to unlock the meaning within unstructured healthcare narrative in an EMR and convert it into structured, actionable data. Perhaps the best way to understand the process of accessing narrative to improve the quality of care is by comparing it to an Internet search engine like Google.
A few years ago, typing "what time is it in Manhattan" into Google's free text search bar would return a few website links. To find out the time in Manhattan, you had to click on one of the links. Type the same question today, however, and Google simply answers, "It is 4:17 p.m." It is almost as though the system understands the meaning behind the question and the desired answer. It doesn't, of course; this "understanding" is accomplished through massive quantities of data, powerful computing tools and clever algorithms.
CNLP is fueled by similarly inventive algorithms that can synthesize and process clinical narrative information into something usable. It can help summarize a patient record, drive clinical decision support, or enable the identification of specific cohorts of patients for clinical trials or other uses.
To illustrate how patients benefit when the clinical narrative is accessible as a result of CNLP, consider another common, high-risk condition: osteoporosis. Physicians know that one of the first effects of osteoporosis can be small vertebral wedge fractures that patients often are unaware they have. As time passes and the osteoporosis worsens, these patients are at greater risk for a hip fracture — a clinically huge and financially costly event.
If a physician were alerted to data indicating a patient had wedge fractures, the provider could intervene with treatment to try and prevent the more serious hip fracture. The problem is that when patients undergo imaging investigations for conditions unassociated with osteoporosis, any discovery of the presence of a wedge fracture is typically recorded in the radiology report as an "incidental finding" via free text.
Because such incidental findings are captured as unstructured data, currently the information generally is not visible to anyone other than the physician who reads the original radiology report. Furthermore, since the incidental finding is not relevant to the condition under investigation, no preventative actions are taken. However, if CNLP is embedded within an EMR, all the information captured in free text — including incidental findings on radiology reports — can become discoverable and usable.
Making EMRs and patient care smarter
Finding osteoporosis patients with wedge fractures and using ejection fraction to identify CHF patients are just two of the numerous examples of how accessing rich data currently contained in the clinical narrative can positively influence patient care and a healthcare organization's bottom line. Retrieving information that currently sits as unsearchable text gives physicians an opportunity to expand the practice of preventive medicine at both the individual and patient population levels. By more easily identifying at-risk patients and reducing readmissions caregivers will no longer simply react to one acute episode after another.
In order to get the most out of an EMR in the era of value-based care, healthcare organizations must achieve as much benefit as they can from existing transactional data, whether it is structured or unstructured. Just as smartphones have become indispensable tools for making daily life easier, CNLP can make it easier for providers to use all the information in an EMR to improve efficiencies, enrich care delivery and strengthen patient outcomes.
Chris Tackaberry is co-founder & CEO of Clinithink.
More articles on health IT:
EHR incentives may have been ineffective, unnecessary, report argues
EMR transfer may complicate UPMC-Highmark split
September's 10 most-read health IT stories