Data is king – no doubt about it. It’s the consummate measuring stick. It’s the beating heart of every healthcare organization on the planet. When it’s in all the places it’s supposed to be, data can present the ultimate gauge of a patient’s health. And, when collected with other carefully processed and organized patient data, it provides a powerful report card for the healthcare business itself.
On all fronts, data is indeed the stuff of life or death!
But what happens when data goes missing? Can key health indicators still be derived? Or does everything stop dead in its tracks?
The Dilemma of Missing Data
Missing data in healthcare can also be interpreted as missing values. Defined, missing data is the “data value that is not stored for a variable in the observation of interest.” In English, that could mean a variety of things:
- the data was not available
- the data was not applicable
- the event from which data was to be derived did not happen
- the data entry person didn’t know the right value
- the data entry person skipped the question entirely
Professional data-miners claim that missing data can be categorized four ways:
- Structurally missing data – data that’s missing for a usually logical reason – typically no entry because it’s not applicable for this sample/patient
- Missing completely at random – missing data that’s entirely unrelated to any other measured variable
- Missing at random – missing data that is unrelated to the missing values but may be related to another measured variable
- Missing not at random – where there is a relationship between missing data and its missing values
Before you get too introspective over which bucket your missing data might fall into, one fact remains: you don’t want missing data.
SPH Analytics Product Manager Walter Maycock, who has worked a few missing-data issues with clients over the years, says, “In most cases, missing data is likely lost data. Perhaps there was a connection issue or ‘hiccup’ while data was being processed. And typically, there are reasonable steps we can take to retrieve the missing data.
“But in some instances,” Maycock adds, “the data is never to be retrieved. Efforts to retrieve it are not justified.”
Sometimes, data that might be considered “missing” may not be missing at all – just in the incorrect location. “On a few occasions,” Maycock adds, “we’ve seen people record the same data element type in multiple areas within their EMR. For the purposes of consistency, data is only extracted from one location within the EMR. The data that was recorded in other areas within the EMR would then appear to be ‘missing’ within quality and population health tracking applications like ours because it was not extracted.”
Prevalence of missing data fortunately appears to be on the decline of late, but as recently as 2015 it was still showing up in top 10 industry hazards lists.
Eliminating Your Missing Data Problem
If left unaddressed, missing data will most assuredly grow to be a significant problem. To best tackle the issue of missing data is to ensure it doesn’t happen in the first place. Maycock suggests the key ingredient in the total elimination of missing data is consistency in data documentation and data entry.
This proactive solution continues to pay dividends when SPH implements its Population Care quality and population health application with its clients. Its primary drivers are:
- the data elements being recorded in the same location within the EMR
- the same content or narrative being used consistently
- ensuring that filters are thoroughly understood by the client, and that they are applied correctly when data is filtered prior to transmission
The Problem of Unstructured Data
To hear some members of the healthcare community tell it, unstructured data is not a problem at all. Patient data that falls under the “unstructured” heading are often physicians’ notes (i.e. SOAP notes), scanned documents, digitized photos, x-rays, audio, et al.
It’s a good argument. Unstructured data indeed provides important contributions to a patient’s record of health, albeit more nuanced and freestyle than data captured under the more rigid “structured” configuration.
And there’s certainly a lot of it out there – nearly 80% of clinical information in EHRs is considered “unstructured” and in no shape to be incorporated into a measurable, scalable format like that required by most quality measure reports.
Which presents quite the daunting challenge: how unstructured data can best be extracted, then measured, and subsequently turned into actionable information. In other words, turning unstructured into structured. Suffice it to say, it’s easier said than done!
Making the Unstructured-to-Structured Transformation
Not to diminish the value of unstructured patient health data, it’s simply too difficult to measure in its “as is” state. And since quality measures account for so much in today’s value-based healthcare, providers and plans want to document as much progress and improvement as possible as their work is gauged. If unstructured data can assist in illustrating that forward momentum, it needs to be exploited. And so should the boost in quality scores that comes with it!
For its Population Care client organizations that could benefit most, SPH has built an advanced algorithm that can accurately capture pertinent data from various unstructured sources, placing it in a format that is scalable and measurable, and meaning that the provider isn’t forced to change his or her documentation style and become less efficient. From that point, the right boxes can be checked, the right clinical codes aligned and confirmed, and the right blanks filled in to both populate an EHR and to optimize a more thorough quality report.
Since unstructured data is, by its nature, all over the place, its capture and process into a formal structured narrative can be a challenge. But it’s one that SPH’s Population Care team is up for.
SPH’s proprietary data collection process, Three-Facet Data Mining, is a largely intuitive process that, once the data has been integrated into a measurable report, can give healthcare organizations insight and 20/20 visibility into their quality measure results so proactive steps can be made for improved performance for MIPS submission or payer reporting.
Is moving entirely to structured data the way to go?
For most companies in the SPH orbit, yes – 100% structured data is the brass ring. But it’s not for everybody.
SPH’s Walter Maycock adds, “Whether you go all-out on structured data really depends on your company’s own goals and strategy. We’ve found a handful of our clients get value from gleaning insights that only unstructured data can offer, most particularly on a one-physician-to-one-patient basis. Some of those health details simply don’t fit within the constraints of a formal and measurable environment.”
But the numbers clearly show that among SPH’s wide client base, the benefits and value of structured data outweigh those of the unstructured variety. Maycock continues, “It’s the cleanliness and consistency of structured data that make it so efficient to work with. It’s orderly, clear-cut, and much more reliable than the alternative.
“And most importantly,” Maycock concludes, “structured data presents a far more accurate barometer of a healthcare organization’s work, its ability to make a positive difference in the lives of its patients, and the group’s own contribution to the healthcare community.” And the accompanying (and improved) quality measure scores would certainly be the indicator of that!
How’s your data? Structured? Unstructured? MISSING? We can help! Send us an email or call us at 866-460-5681.