Deprecated: The PSR-0 `Requests_...` class names in the Request library are deprecated. Switch to the PSR-4 `WpOrg\Requests\...` class names at your earliest convenience. in /var/www/wp-includes/class-requests.php on line 24
NLP Archives - SyTrue

SyTrue has been acquired by ClaimLogiq

The same NLP Technology created with you in mind, just a new home!

Please wait 10 You will be redirected shortly.
Visit ClaimLogiq

You see two prominent terms in many healthcare headlines these days: “big data” and “smart data.”

They’re familiar, yet we don’t quite know what they mean. Let’s parse them to answer one question: How do we use data optimally and “smartly” to help providers and patients in healthcare settings?

Is all data easily “convertible” into “smart data”? Yes and no. Smart data requires “extrusion” and special steps, as we’ll see below.

Take the Apple Watch, which offers you lots of data. Does it “do” smart data? As a tool (debuting in 2015), it quizzes the wearer, monitors heart rate and steps walked. Via sensors, it tracks pulse and blood pressure. With other apps it may monitor blood glucose levels in diabetic kids – which puts it on a path to providing smart data.

A useful watch, big potential, but no “smart data device.” Only when your GP can use your watch to track critical “diagnostics” and help you reshape your health behavior with many “data streams” – that’s when we get smart data. Healthy behavior is key (forget for now how we get patients engaged in it). This won’t “score” on the smart data scale.

Google Glasses in the ER
Google Glasses in the ER

A better candidate: Google Glass, now being piloted in major hospitals, and producing good results. As Beth-Israel Deaconess CIO John Halamka reported earlier, it’s in BID’s Emergency Department, and excels in data gathering at time-critical moments. Example: A patient with a profound brain bleed – on blood thinner plus having “some” allergy to blood pressure medications – needs immediate attention. The Google Glass Wearable Intelligence data-feed lets physicians “see” the patient’s medication regimen quickly, offers his allergy problem list yet lets them keep eye-contact and act in minutes. No logging-in to EHR records or extra phone calls.

So it “scores” on the “smart data” spectrum. It accesses and confirms vital clinical information, which drives crucial decisions in that ER moment. As CIO John Halamka sums it up in his blog: The tool offers “contextually-relevant data and decision support wisdom” – a fine starting point for characterizing smart data.

So Google Glass’s is a clear advance as “smart data” — allowing quick, richly-sourced decision making in critical moments. But there’s more to this picture.

While media stories and marketing of big data continue, it’s clear we’ve entered a new phase, a time when “smart data” supersedes “big data” in healthcare.

Figure 1 – GARTNER on the “BIG DATA” Hype Cycle (July 2013): “Big data” has peaked in “inflated expectations; it’s ready to plunge into the “trough of disillusionment.”


Big data is everywhere, but “smart data” is harder to find. We need it because it provides vital “wisdom” for complex decisionmaking – in population health, in ways to enhance radiology results, in improving revenue cycle management, operations and workflow.

With smart data, we flex the right mental muscles and tackle tough questions. Example: Which hospital patient, due for discharge, is a candidate for readmission in 30 days? “Smart data” can answer this, saving the hospital significant ACA penalties in the process.



Penalties for (within-30-day) hospital readmissions (by ACA law):

Average 2014 cost per penalized hospital: $102,022.


We have new analytic tools to process all data, but many are ill-suited to healthcare’s needs. The volume of data alone is an issue; there’s also the variety of data to “extrude.” Google’s Eric Schmidt tells us that from early history onward, humans created (in total) 5 exabytes of data. But, he adds, “We produce five exabytes every two days [now]…and the pace is accelerating.” Even Google’s storage systems are challenged: They have 3 separate levels of complete backup – every single day.

So as data volumes grow exponentially, we need more “smart data” – context-rich, “decision”-driving information to help healthcare providers answer big questions.

But “processing” the expanding volume of data, and extruding the right “decision driving” wisdom – smart data — from it is the crucial hurdle.

Consider the many data sources you need to convert “regular” data into “smart data wisdom.” On average, all hospital’s may have 180 systems either monitoring or collecting information – some not linked to others.

If you’re a data scientist looking to extrude the smart data, you face Emergency Department records, physician lists and orders (admitting, attending), observation records, radiologist/lab reports, EKG reports, consent forms, progress notes, discharge notes, discharge summaries, continuity of care records, pathology reports, referral notes, registration forms, nursing and physician notes, EHR records (just a partial list of sources). These arrive in different codes – at “source data” level – and formats.

Interoperability (usefulness) of records from elsewhere is another layer of issues. And you have “unstructured” records (notes, transcriptions) offering extra problems.

This is a “dirty” data picture — until it is Parsed, Cleaned and Refined. That means identifying the data type, processing it to normalize it (despite the varied formats and native coding used). For smart data purposes, you need to make the information to be plain language searchable (natural language processing steps), smart in spotting mistakes, and “decision”-ready for the many providers hungry for it. It’s a treasure trove, but at the outset, it’s just a big bundle of “ready to clean” bits.

At SyTrue, we saw some years ago that data variety, data size and interoperability would be dilemmas years ago. We’ve designed our data “extrusion” process so that we can take any data and convert it into a consolidated CDA – making it searchable and usable (via semantic interoperability). Until clinical data is encoded at the point of care, we won’t have true interoperability through non-“smart” systems such as EMRs and EHRs.

For smart data purposes, EHR lack the clinical intelligence wisdom that gives healthcare providers deep information on patients and patient groups. SyTrue’s smart data platform, in addition, actually “sits” on top of an EMR/EHR, but below the electronic health information exchange (HIE) a provider or hospital uses. Our platform offers interoperability with other systems and smart, “decision ready wisdom” on demand.



The problem has been until recently the only place analytics was used is billing, not CLINICAL INTELLIGENCE. There has been no incentive until recently to look at clinical analytics. The systems used to support reimbursement (EMR) are not good at clinical data capture. So providers only have access to retrospective claims data. Then each provider attempts to put their own spin on stale information. This will ultimately limit their effectiveness to drive outcomes and clinical change. If it was useful the United States would not be at the BOTTOM of the quality bucket for rich country’s.

America’s health quality is more like a poor to middle country than a rich country. So I have to ask what value does this extraordinary high priced of care bring?

IMO…very little but profits for a small few while the majority of the population suffers. If Claims analytics worked we wouldn’t have 40 wrong side surgeries per week and medication adverse events as the 3rd leading cause of death. If you looking at vendors who only know the INCOMPLETE view of claims, your putting your ORGANIZATION and its PATIENTS at real RISK.

If you REALLY want to do MEANINGFUL ANALYTICS than you have to look beyond claims to “dirty” UNSTRUCTURED clinical data at the point of documentation. This is where SyTrue comes into play, we are the leading solution in the nation who addresses this problem. Our ability to provide Encounter Based Analytics is UNSURPASSED. You MUST have the clinical narrative or will have INCOMPLETE and possibly INACCURATE view of your population. So if you want accurate, population based Analytics you need SyTrue and its Clinical NLP, Semantic Search, Terminology Server and real time Business Intelligence and Encounter Based Analytics to achieve a Data Driven Approach to quality improvement.

Skip to content