The 65-radiologist Austin Radiological Association is well known for its robust information technology deployment throughout central Texas. Nonetheless, the practice’s billing department, which billed close to a million procedures last year, was awash in the same paper that plagues smaller, less sophisticated practices. Enter Laura Casey, business office director, who partnered with CIO R. Todd Thomas, and a team that included representatives from both departments, to design and write software architecture for an automated billing process that eliminated processing paper.

Using an HL-7 interface engine, a document scanning application that can retrieve text from a document and export that data into a database, and an in-house software engineer, Casey, Thomas, and team devised the new process over a period of a year. And after an initial two-month testing period, the results are very promising. Review of the initial automated charge transactions indicates the practice is within one percentage point of the previous year’s collection rate, claims and patient statements are accurate, and the billing staff has been reduced by 75%.

To fully appreciate the achievement of the new system, it is important to review the former process. “The old process was all about managing our paper radiology reports and patient demographic sheets,” Casey explains. “We would touch each radiology report approximately six times to manually key demographics and charge transactions into the billing system. We organized the paper into batches by date of service and facility. Understanding that an entire hospital’s date of service does not arrive to our facility on the same day, but trickles in over a period of a week or so, we had to organize the paper into batches by date of service and facility and then wait until we thought we had everything for that day. Our next step, to ensure proper 76 and 77 modifier coding, was to alphabetize the batch by patient name.

“The sorted and alphabetized batch was then reviewed by a coder and coded the old-fashioned way; taking the pen to the paper. Once the coder completed a batch it would head back to the same group of people who sorted and alphabetized it. They would log onto a hospital browser, find the demographics appropriate for that exam, and create more paper by printing the face sheet from the browser. Finally the batch would move on to a person who would enter the data into our billing system. From that point the billing system ensured the claim was on its way to the payor.”

“It took 20 people to create charges and update accounts last year,” continues Casey. “Another challenge rested with not knowing what was lost or missing. It was really hard to know what we were missing. We created a missing process, where we compared our billing data quarterly to the hospital’s data to learn which reports had and fallen prey to paper jams, empty toner cartridges, or lost faxes. Our missing process unearthed many lost pieces of paper, and it was effective. However, when we received our new-found missing reports, we had to process the paper through all that the steps I just described.”

Clearly, the manual system was cumbersome, with many potential failure points along the way. “The real pain point for us was getting the charges into the system,” Casey notes. “We knew we’d have to continue coding them manually, and chose not to automate the coding process with coding software, because the larger cost burden resided with all of the people sorting the paper, printing the demographics and then finally performing the data entry.

ARA’s goal was to reduce all of that effort, and to do it in such a way that would be scalable. The practice had served Austin since 1954, and wanted to maintain the same high level of customer service as it grows along with the thriving city. So, the question, according to Casey, was this: “How do we bill more accurately and timely without incrementally adding the most expensive cost to the equation: people?”

When discussions began internally about creating an informatics-based fix, Thomas happened to be looking at an HL-7 interface for another project he was working on. So Thomas assigned one of three staff software developers, and provided him with the HL-7 interface engine and another product called Captiva, an application owned by EMC that is able to scan text the off documents and faxes and export that data into a data base.

“Technically, we had to marry the two pieces of the puzzle,” Thomas explained. “We had to figure out where we were getting the demographic information from, we had to scan in the text from the radiology reports, and then write logic that was matching the demographics to the diagnostic report text. We send all of that to a coder through an application that we wrote in-house so that they could then, very easily, go through and code the diagnostic reports. Then we take all of that information out of the database, push it through our interface engine, convert it to an HL-7 message, and upload it into our billing system.”

ARA chose to carve out of the automated process persons under the age of 90 days, because newborns receive a lot of services, and in the case of twins or triplets the opportunity for mistakes rises dramatically. “Now, instead of doing 100% of everything manually, the process focuses our work on the exceptions. If there is something missing or odd, then a human being gets involved,” Casey notes.

As the new process was rolled out, some adjustments have been made based on unforeseen issues. “When you put everything together and start running batches through the system, you realize, we need exceptions for this,” Thomas explains. “We had to re-set our insurance codes lookup to look for sample A, sample B, sample C, for instance, because the hospital may use different codes for the same insurance company, or different addresses for the same insurance company, or zip codes aren’t properly loaded.”

Having endured the process of building numerous interfaces between the practice and hospital image and information management systems five years ago, ARA decided to take the project in-house. “Getting an interface built between a hospital and a physician billing system,” Thomas says, “the hospitals could never send you everything perfectly formatted for a computerized system. You were constantly relying on a vendor to make this change and that change, and pretty soon the interface they’ve sold you has been completely rewritten. So after we went through that process six, seven years ago, over and over again, we just decided last year that we were going to do it ourselves.”

ARA’s new automated billing system is in what Casey calls a limited-live mode at the moment, and she is proceeding with caution. The first batch of 1276 charges that went out in June is about one percentage point away from being completely paid and Casey has noted only minor data and logic points in need of fine-tuning. “We are moving judiciously toward 100% live production but want to be sure that we don’t adversely affect the company or our patients,” Casey explains. ““It’s not only the revenue but the integrity of our data, which translates to the integrity of our patient’s claims and bills. Although we couldn’t have asked for a finer more talented team to successfully create such automation, the process is new and I’d rather be safe than sorry. ”

Aside from the ability to reduce staffing levels the system offers many potential benefits. Reduction of errors and expeditious posting of revenues are two large benefits. “I estimate 90% to 95% of our charges are going to flow automatically once they are coded,” Casey says. “For the long haul, we hope to manage only 5-6% of the charge volume with human intervention. The opportunity to fall behind and the opportunity to let things stack- up is a much smaller opportunity now. We will become very current very rapidly, and then we have a greater ability to remain in that position with no additional staff even if our volume increases. By the end of the third quarter the beginning of the fourth quarter we will be sending out most of our charges via the automated process.”