Earlier this year, Council for Affordable Quality Healthcare (CAQH) released their 2020 Index which measures the progress of adopting automation for a variety of administrative processes. The upshot is that billions of dollars are being saved by converting manual processes and the data they use into electronic transactions and 2019 increased savings by $20 billion to a total of $122 billion. Yes, that is $122 billion; a large number to anyone not named Jeff Bezos.
But even as transactions such as claims submission have a high degree of automation, with estimates over 90%, there are processes that represent a significant amount of cost in both terms of time and money. Take, for instance, the problem with attachments.
The overall volume of attachments involved within medical transactions is relatively low at less than 1% of all transactions but remains a significant headache. The reason is that, while not an administrative transaction in and of themselves, attachments are a critical component of many processes including prior authorization, claims submission, referrals, audits, and more. And when they are involved, progress often grinds to a halt.
To understand the issue, it’s important to address the subject from a data perspective. Attachments are most often medical records such as labs, physician orders, and progress notes. Even with the adoption of electronic medical records systems, which removes the paperwork, the information contained within records is still largely what we call “unstructured data”. It is considered unstructured because much of what doctors and other medical staff type into the system is in the form of text that is not labeled. This means that the ability to easily translate information contained within a system, such as a diagnosis, to another format is extremely limited. And when it comes time to share the information stored in record systems, regardless of whether the system is of the old paper-based colored file variety or a modern web-based solution, processes tend to resemble something from the last century.
CAQH measures automation of attachments for two processes: claims submission and prior authorization, and, believe it or not, the primary means for these documents to be shared is via fax machine or mail. In a separate report, CAQH estimated that over 80% of attachments arrive in this manner. This means that providers often print out records in order to share them, turning digital information (even if unstructured) into images or paper, requiring a separate document handling and data entry process to manage them on the receiving end. One regional health plan that participated in the study revealed that it manages the equivalent of 20 full-time staff just to process attachments it receives.
So even at a relatively low transaction volume, the cost to manage attachments is significant. There are a number of factors involved that conspire to keep the attachment disorder a problem.
Data Complexity. Even if born-digital, medical records don’t always remain in a structured format. Electronic medical records systems may provide form-based data entry for common data such as service date and patient name, but a significant amount of data within records consists of unstructured text making the process of finding specific information a time-consuming and complex process.
Lack of Standardization. Unlike other transactions that have strong support for electronic standards from the Department of Health and Human Services, of which CMS is a part, medical records continue to be a stubborn thorn in the side of medical transactions. A big part of this is the nature of information as explained above. But another problem is the various agendas of EMR/HER vendors that aren’t all that interested in enabling easy conversion of data from one system to another – there is an incentive to create “vendor lock-in”.
Inability to Start 100% Digital. Medical records, even those several years old, represent a valuable information repository. The value of older data is different from the value of more ephemeral transactions such as prior authorizations where, once completed, largely don’t represent useful information. This means that the ability to extract full value from a fully digital records system requires a significant amount of expense converting older paper-based records into a specific format.
Transactions that are electronic, aren’t always that way from start to finish. As evidenced in the CAQH studies, the most common method of exchange is via fax or mail. Once a digital signal is converted to analog, the ability to easily use the information is significantly hindered. When it comes to document-based information, it means scanning paper and converting images to text via lossy OCR.
Portals are a start but not the finish line. Many “digital solutions” are actually only delivering a partial answer to creating a digital transaction preferring to offload the problems with attachments onto human staff. Providers may get a slick web-based application into which to enter their PA request, but they only replace the fax machine with a process of extracting medical records, scanning them, and then importing them into the payer system. And then the payers must make sense of the files and perform arduous data entry.
There is Hope
I am not going to use a hackneyed phrase like “machine learning to the rescue” even though it has a significant part to play. But there is hope in the form of machine learning-based document automation. Yet even the most sophisticated deep learning systems will only work if the workflows associated with document handling also change. Next time, we’ll cover how processes can be adapted to use technology all the way through but without significant disruption.
Be sure to join us at Claims Innovation USA where industry veteran Mike Hurley of BRYJ Consulting will discuss the art of dealing with attachment disorder.