- the inconvenience of data entry stops tools being used by doctors
- computer-assisted physician documentation (CAPD) can alleviate the problem
- natural language understanding (NLU) converts sentences into codes
- One of the most exciting promises of Big Data combined with advanced technologies like NLU is its ability to “bootstrap” — to identify important missing data and either grab it from another source, infer it from existing data, or prompt the clinician to add it during the normal course of documentation.
Source: http://www.wired.com/insights/2013/12/big-data-insights-without-big-data-entry/
Big Data Insights Without Big Data Entry
- BY DR. JONATHAN HANDLER, M*MODAL
- 12.23.13
- 2:03 PM
Image: openexhibits/Flickr
My brand new Xbox One has a new feature — its Kinect camera recognizes users’ faces and automatically logs them in. The new feature saves me nine clicks and 45 seconds compared to my old Xbox 360. Microsoft correctly recognized that people are too busy for their videogames to turn them into data entry clerks.
It’s the same thing for doctors. Doctors can use Big Data algorithms to help provide better patient care, but those algorithms are very data hungry. Who will enter all that data? Patients already wait weeks to be seen and then sit in the waiting room despite an appointment. The last thing we want is doctors doing even more data entry that further delays care. At the same time, we want computers to provide great decision support to doctors.
Can we realize the benefits of Big Data without suffering the pain of Big Data entry? Using modern technologies, this may be possible.
In 1998, my colleagues and I had built some of the earliest online decision support tools, and within a few years more than 50 were freely available. Although very useful, I used those tools only occasionally in my clinical practice because most required me to enter lots of data before providing any help in return. For example, one score for predicting the likelihood a patient will die requires the user to enter 17 pieces of data, two of which are formulas that must be calculated. Big Data can speed up the development of decision support tools and create more accurate algorithms.
However, those algorithms require lots of data. For example, an algorithm for diagnosing heart attacks was developed using 156 data elements, and the final model used 40 data elements. Entering all that data is a lot to ask from physicians trying to manage an already overcrowded emergency department.
Over the last decade, it has been well documented that tools demanding significant data entry were felt to make doctors less efficient and were seen as less useful. Not surprisingly, tools requiring doctors to do lots of data entry were used less often. Although automatically populating these tools with existing data from the Electronic Health Record (EHR) seems the obvious answer, it has real challenges.
Sometimes it is not clear which data to enter. For example, there may be multiple blood pressures documented and it may not be obvious to the computer which one should be used. In other cases, the needed inputs might not be recorded exactly in the right format or with the right details, or the inputs may be completely missing.
New technologies might alleviate the problem. Computer-assisted physician documentation (CAPD) uses real-time natural language understanding (NLU) to convert the clinician’s sentences into computer-readable codes.
Those codes automatically populate decision support algorithms executed by rules engines, and seamlessly display the result. Caregivers are no longer forced to re-document the same information in checkboxes and manually run the tools. The rules engines automatically identify when a required input is missing, and can immediately prompt the clinician to add it if warranted. This enables real-time decision support, so that clinicians can provide better care.
CAPD becomes even more interesting when the rules engines start doing even more work on behalf of the clinician. For example, in addition to auto-populating decision support rules, the system might even save time by auto-generating some of the doctor’s notes, auto-generating billing and regulatory reporting codes, and automatically building a provisional set of orders based on the doctor’s notes.
One of the most exciting promises of Big Data combined with advanced technologies like NLU is its ability to “bootstrap” — to identify important missing data and either grab it from another source, infer it from existing data, or prompt the clinician to add it during the normal course of documentation. With these technologies in place, we can reap the value of Big Data without paying the heavy cost of Big Data entry. The result will be better and faster healthcare for all.
Dr. Jonathan Handler is the Chief Medical Information Officer at M*Modal.
