Be aware that speech recognition software can introduce errors to the medical record. Errors can threaten patient safety and increase liability.
- Speech-to-text is sometimes offered to make up for the increased work required with electronic records.
- Mistakes occur most often in long narrative sections of the record.
- Names, locations, and similar-sounding words are most problematic.
Many physicians rely on speech recognition software to speed the process of entering information into the electronic medical record, but one doctor says risk managers should be aware the software is imperfect and can introduce potentially harmful errors.
Par Bolina, MD, an internist in New York City and chief innovation officer with healthcare consulting company IKS Health, has used common software dictation programs in the past, dictating notes for about 25 patients a day. Some hospitals and healthcare systems offer the use of speech-to-text software when introducing an electronic medical record, as a response to concerns the electronic record is burdensome and time consuming for physicians.
The dictation software can be much faster than the traditional method of dictating notes and waiting for a transcriptionist to enter the notes in the record, Bolina says, but it is far from perfect.
Rather than having a knowledgeable transcriptionist edit and properly structure the physician’s comments in the record, some speech-to-text software merely records every utterance and enters it into the record, Bolina notes.
“It actually expects you to do all of the editing, choosing a comma or semicolon, capitalizing words and finding drug names,” Bolina says. “There are more sophisticated and expensive software versions designed for physicians, and the accuracy of capturing names of diseases, drug names, and medical terminology is much better.”
Even with that advanced medical software, the physician must still interject every punctuation and sometimes spell challenging words, he notes. Bolina found that entering accurate data into the record in this way was more taxing than handwritten notes or the old way of simply dictating for a transcriptionist who would take care of the details.
“Cumulatively through the day, you really look forward to your lunch break or anything else where you could do something other than dictate these speech-to-text notes because the level of concentration was very high. It’s intense,” Bolina says. “Certainly by the end of the afternoon’s batch of patients, you are somewhat drained. There is a cost of effort that has to be factored in when the physician day in and day out is trying to put together a thoughtful and concise summary of the critical elements of the visit. It’s more than one realizes in terms of time consumption and brain power.”
Additionally, the physician has to be cautious about background noise or someone stopping by to ask a question. The recording must be rewound if necessary, taking more time and more effort.
“You end up being not only the composer of the note, but also the final editor. You have to create the note, which is taxing and time-consuming, but you also have to go through it to make sure no sounds came in and altered the document,” Bolina says. “If you’re not that careful and you’re in a circumstance where that note is being audited or reviewed, it is difficult to defend an overt error. How do you explain to an auditor or a jury why a string of nonsensical words was entered into this important document?”
Mistakes like that will imply to some people that the physician was careless, at least in creating the document and potentially with the medical care in question, he says.
“You have to go through an extra level of scrutiny to make sure that the document is complete and says what you really want it to say,” Bolina says. “That is why adoption of speech-to-text software probably won’t penetrate more than 20% of physicians. You have to be pretty adept at being someone who can think out loud, punctuate pretty well, have an eye for detail, and do it correctly day in and day out.”
Even without a majority of physicians using speech-to-text software, the risk of errors should concern a risk manager, Bolina says. In fact, he says, having a minority of physicians using the software may mean it is less likely the physician practice or hospital will address the potential for errors in a meaningful way.
Errors from dictation software are most likely to occur in the two parts of the medical record with the most significant narratives, Bolina says. The first is in the history of the present illness, which documents why the patient is being seen or the reason for the illness. The second is toward the end, where the physician outlines the plan for treatment.
Mistakes are more likely to be overlooked in the history of the present illness because the physician is summarizing comments from the patient or answers to questions rather than actively developing his or her own thoughts and speaking them, Bolina says.
The most common types of errors are substitutions of words that sound similar, Bolina says.
“If I’m moving quickly, I may not notice,” he says. “It’s like if you dictate an email on your phone, you’ll notice that some words come out not at all like what you had in mind. It may be two words instead of one, or it may be a completely different word that makes no sense in this context. That’s no different than when we use speech-to-text in clinical practice.”
The software does a surprisingly good job with the names of diseases and medications because the manufacturers can build in those standard terms, Bolina notes. But the software has more trouble with the names of physicians and locations, for instance.
Some physicians are moving now to more advanced technology and services that are a sort of hybrid between the old-school technique and the more recent speech-to-text. In some cases, the physician can dictate the note with speech-to-text software and that information is reviewed by a physician at the end of the day, corrected as needed, and sent back to the dictating physician for final approval and entry into the record.
“There is a way to make speech-to-text work for physicians and maintain a reliable electronic record, and allow the physician to focus on patient care and clinical decisions instead of so much time on the note,” Bolina says. “The key is to know that simply providing speech-to-text capability is not necessarily the solution to the work required with an electronic medical record, and you need to consider how this is affecting the reliability of your records.”
- Par Bolina, MD, Chief Innovation Officer, IKS Health, New York City. Telephone: (323) 417-6565. Email: email@example.com.