The Switch to Speech Recognition

Communication, set goals and good service make the difference

You don’t have to search too long or hard to find stories about failed speech recognition implementations. However, the tide continues to turn toward the technology. Make a new year’s resolution to focus on adequate preparation, your expectations and the service you need, and the technology can offer dramatically improved turnaround time, better patient care and more satisfied clinicians.

Community Health Systems, Inc. (CHS) based in Brentwood, Tennessee, is the leading operator of general acute care hospitals in non-urban markets throughout the United States, generally in areas with low managed care. CHS operates 73 hospitals in 22 states.

Northeast Regional Medical Center (NRMC) in Kirksville, MO, is one of 73 hospitals in 22 states operated by Community Health Systems Inc, based in Brentwood, Tenn. The facility implemented SpeechQ speech recognition from MedQuist in September 2005.

Aside from wanted to improve report turnaround time, the hospital had gone from using the same local radiology reading group for more than 40 years to a string of locums, says Julie Atchley, SpeechQ administrator. That led to new dialogues from the reading radiologists that were harder to decipher. “We thought speech could aid with that,” she says.

Hospital administration decided on SpeechQ and MedQuist sent in a team to set up the system and train the users. NRMC specifically implemented speech recognition for radiology but at any given time, only two radiologists were in house. Atchley and a coworker learned the system so that they could catch the radiologists on the fly and train them. She says that’s all it took to get them using the system proficiently.

One of the biggest proponents of SpeechQ at NRMC was a veteran doctor who had been with the original reading group. He was used to using a handheld device and was leary of the new system. However, he quickly realized the advantages of SpeechQ and spread the word among the other radiologists. “He is very positive about SpeechQ,” she says. “I thought he would have the biggest hesitation but he latched right on.” Even though some of the locums complained about creating their own reports and self-editing, they’ve seen this doctor efficiently producing reports which has gone a long way to sell the benefits. Atchley also made every effort to sit down with the complainers to help them learn the system and figure out how to get it to do the tasks they want. “It was a really great process.” Between her efforts and all the perks of the system, “basically we’ve won over all the physicians,” she says.

Within two weeks of implementation, “we were dedicated to SpeechQ,” she says. After a couple weeks, she was able to take away all of the old handheld dictation devices. Now, report turnaround time is less than 30 minutes during the week and averages seven to eight hours when considering the weekend as well.

Another CHS hospital went through a speech recognition implementation at the same time as NRMC, Atchley says, but they failed. She attributes the difference to her facility’s administration knowing exactly what they wanted to get out of the system.

NRMC eventually signed with a new radiology reading group. The doctors like SpeechQ so much that they bought a system for their practice.


Ready for prime time

Many clinicians tracked speech recognition waiting for the technology to mature before implementing it. After watching it develop over several years, Alan Hecht, MD, chairman of the radiology department at Mt. Sinai Hospital in Chicago, IL, says he never thought it had come far enough along to be usable. After testing PowerScribe he quickly changed his mind. That was two years ago. Now, about 15 radiologists are using PowerScribe and experiencing turnaround times of hours compared with the days a report took before implementing the tool.

The hospital installed PACS and then added PowerScribe a few months later. Although considered one big project, the staff knew they wanted to give the PACS a little time to get up and running smoothly before introducing the physicians to speech recognition.

One of the biggest concerns was report turnaround time, especially for emergency patients. As a level one trauma center, every second counts so turnaround time is tracked every week. With speech recognition, 80 to 90 percent of ED cases have dictation available in the medical record within one hour from the time the patient entered the room for an exam.

Radiologists signed on to the speech system one at a time. Everyone was very nervous in the beginning, reports Hecht. Especially concerned were those with strong accents but they have done as well with the technology as anyone else, he says. “After a few months, [the radiologists] came to me and said they didn’t expect it to be this good,” he says. As the physicians were phased in, the transcriptionists were phased out. At first, the doctors could send reports in for editing but that gradually decreased and now they all self-edit.

“It’s been way above our expectations,” says Hecht, and the numerous benefits were well worth the anxiety. The system learns each user’s voice over time, so it gets better and better. And, the easy use of templates saves a lot of time.

Support from PowerScribe was on site for a week and the facility’s PACS administrator had enough training to offer ongoing support.

For radiology, speech recognition has gone beyond the tipping point, Hecht says. In fact, Mt. Sinai’s pathology and cardiology staffs have been told to follow the radiology model because the turnaround time is so fast. “They want the entire hospital to be the same,” eh says.

The hospital has hosted several potential speech users who want to see it in action. “They all come away very impressed,” says Hecht. Seeing and using the system in person helps alleviate the anxiety and, the improved turnaround time is a big sell as well. “All of radiology is going to go to this and all of medicine,” he says. “Anyone [who is] interested in improving turnaround time and quality of care.”


Improving transcription productivity

Consolidated Radiology Complex (CRC) in Caguas, Puerto Rico, serves providers in the southern and northeastern regions of the island. CRC had been using digital dictation from Crescendo but when they upgrading their computer system, they decided to also upgrade to speech recognition. CRC installed several software modules for a customized dictation and transcription solution. The six radiologists use DigiDictate-IP, a dictation interface that lets them dictate directly into their computers in real time. Once a dictation is complete, Crescendo’s iRouter routes it to the SpeechMagic server or the transcriptionist depending on the preferences set.

Unlike many facilities that replace transcription with speech recognition, CRC wanted to keep its transcriptionists, says Marirosi Martinez, transcription manager at CRC. The goal instead was to help them double their output so she wouldn’t have to add to her staff.

In fact, dictation has increased by 40 percent since the beginning of 2005.

The implementation went so well that Martinez says the facility is currently looking into front-end speech recognition. Good communication is the key, she says. “When you have a change process like this one, you need to communicate everything, get all the people involved and let them know what to expect and get prepared for it.” The only way to avoid anxiety is communicating as much as possible, she says.

Once the decision was made, Martinez says it took about three months to get everyone ready to use the system. Despite the distance between Puerto Rico and Crescendo’s Montreal headquarters, she says service was excellent and two of the company’s IT engineers were onsite for the transition.