DRG Coding Advisor: Autocoding brings coding to 21st century and beyond
Myth: It’ll cut jobs; Reality: Coders become experts
Only in a science fiction medical fantasy will the process of autocoding replace hands-on coders in hospital systems and other health care settings.
Autocoding is the process in which a clinician dictates notes that are transcribed and sent to a third-party vendor who has a software system that uses speech recognition software to automatically apply codes to diagnoses and services. The resulting coded information is returned to the health care provider’s coding department.
However, the autocoding process, now in its infancy, is neither simple nor perfect. Even if autocoding greatly improves and becomes a standard technology for health care providers, it still will require expert coders to analyze and interpret the codes before they can be used with confidence for billing and quality improvement purposes, according to an expert on this new frontier of the medical coding industry.
"The best way to describe this is if we really take a look at what’s going on in coding today with the advent of APC codes and all of the compliance issues," says Todd Karner, BSN, MGA, senior sales consultant for SoftMed Systems of Silver Spring, MD. Karner is scheduled to speak about autocoding at the 74th National Convention and Exhibit of the American Health Information Management Association of Washington, DC, held Sept. 21-26 in San Francisco.
Karner says he doesn’t expect autocoding to replace coding staff. Ideally, health care systems will use autocoding to sort out routine coding from the more complex coding work, so coders and HIM professionals can use their time more efficiently, he says.
"Let the autocoders handle the easy things, and let the coding professionals take the time they don’t have right now to really focus on where they are most needed," Karner says.
Karner offers this preview of autocoding developments:
• Pioneering autocoding vendors and software: A-Life Medical Inc. of San Diego, located on the Internet at www.alifemedical.com, is one of the leaders in natural language processing, which is how the magic of autocoding works, Karner says.
"The first generations of autocoding basically look for key words in the documents that are being evaluated so they can see a word fracture that would trigger a number of potential codes," Karner says. "It works differently in that it tries to evaluate words in the context of other words around it, even at the paragraph level."
Therefore, the autocoding system will be able to differentiate between a medical history and other medical chart sections. For instance, if a physician has dictated information about a patient’s history and the software picks up the words "fracture," "femur," and "history," then it knows that this information is a history of patient and not an active problem.
"It also can take words in conjunction with other words in a sentence to get a more intelligent interpretation," Karner says.
This level of computer intelligence is critical to it being used effectively. "The software can rule out that this has not happened, as opposed to what has happened, and it applies logic to the text," Karner explains.
• Catching coding mistakes: Each vendor has a set of flags that gives the user an indication of how the coding session went, Karner says.
"Obviously, a number of these coding sessions are not completely perfect, so the vendors then rate the coding session as far as accuracy," Karner says. "Some give you a numeric rating or flag, and they identify whether the coding session was clean and whether there were problems with it."
At this point, coders will manually review the process to correct the mistakes.
This is where autocoding as a technological efficiency could succeed or fail. If an autocoding system generates code sets that are riddled with errors, requiring hours of coder review, then it may be considered too much trouble and expense to use. However, if an autocoding system achieves an error rate of 10% to 15% or less, then it may be worth the investment, Karner says.
The percentage of errors likely will be directly related to how complex the coding files are. This is where a health care provider can both improve coding accuracy while enhancing the efficient use of staff resources.
"The strategy may not be to have it do all the documents, but if it can do the easy ones and leave the other ones for coders, then it may be a strategy to adopt," he suggests.
• Early users of autocoding: At present, the areas most likely to use autocoding are emergency departments and radiology departments, Karner says.
"Probably the reason for that is they tend to have a fairly well-defined set of patient populations with minimal variation," Karner explains. "And the other piece that lends itself well is that complete documentation is available electronically in those two areas."
For instance, if an emergency department physician dictates ED notes, there’s a greater likelihood that there will be a completed dictation in a single document than if the physician is describing an inpatient stay. The same is true of a radiology note where a chest X-ray will require a single document of interpretation, Karner says.
• Structuring the autocoding document: This typically is a greater problem than the vocabulary issue. Different facilities and different physicians may put important information in different sections of a report, and this can create problems for the autocoding system.
"It’s not uncommon for physicians to mix and match different pieces of a document when they do dictation," Karner says. "If they start to talk about their diagnosis when discussing the hospital piece or treatment, it makes it difficult for the autocoder to work."
A way to prevent this is for an HIM department to show clinicians a single format for dictating patient notes and to follow up on their progress in adhering to that formula.
One way to accomplish this is to post notes with the proper order for dictation, e.g., first: history of presenting illness; second: review of systems; third: impression/diagnosis; fourth: treatment plan. "Give physicians wallet cards with this order," Karner suggests.
While transcriptionists can help out with after-the-fact reformatting, this isn’t the best solution, Karner notes. "It’s better if it happens on the front end because there are so many variations in transcription these days from outside transcription services and voice recognition systems, and they may not offer the same services as in-house transcriptionists."
Which records need human interventions?
• Coding work after autocoding: If an autocoding system were to work ideally, then the HIM department would receive the codes with the flags, and the coders could then begin to clean up the errors. Everything else would be assumed to be accurate.
However, that isn’t the case, and there likely will always be a need for coder intervention, even among the supposedly accurately autocoded documents, Karner says.
"The key is to identify which records need human interventions and which ones don’t," Karner says.
For example, suppose there are 100 records that return from the autocoder. Assume that the hospital system’s physicians dictate very well, so 75% of those records come back as completely clean. In this scenario, you can either send those clean records directly to billing, or, from a quality assurance perspective, there can be a random selection of a percentage of those clean records for review by an HIM professional, Karner says.
"I think this is where integration between autocoding functionality and traditional HIM vendors that support autocoding products are key," Karner says. "Most clinical abstracting products have a workflow that automatically routes the appropriate visit to a coding resource, and I think that having a user definition on how those jobs should be routed becomes essential in the autocoding process."
When autocoded records return to HIM departments, coders will not be coding these items, but will take the results and look for mistakes to correct, including subtle coding changes that will more accurately reflect the intent of standardized coding guidelines.
Subscribe Now for Access
You have reached your article limit for the month. We hope you found our articles both enjoyable and insightful. For information on new subscriptions, product trials, alternative billing arrangements or group and site discounts please call 800-688-2421. We look forward to having you as a long-term member of the Relias Media community.