Lettering in Documentation Quality

At SpeechCheck, we are passionate about high quality radiology documentation. Admittedly that seems a little … weird. But is it really? Ever stop to think how a single letter can impact patient care?


The bulk of the documentation that we audit is produced by radiologists using speech recognition systems. These systems are reasonably good at interpreting dictation. But they are far from 100% accurate. Consider just two of the most common error categories that we see.


    • Measurement mistakes. Why are measurement errors common in speech-recognized text? Because context doesn’t provide sufficient clues for interpretation. Although they are numerically far apart, the words “15” and “50” are phonetically quite similar. And that’s just one example! It’s much easier for a computer to hear the words “there is a blockage in the right aorta” than for it to accurately interpret the dictated measurement of the blockage. Think about it – can you come up with any other phrase that sounds like “blockage in the right aorta?” No. But you can probably come up with any amount of measurements that sound quite similar (15 vs. 50, 16 vs. 60, 3.8 vs. 3.0, and so on). Moreover, physicians tend to be less confident when dictating numerical values, further diminishing the chance of increased accuracy on the recognition system’s first pass.


  • Sound-alike antonyms. Okay, I just made up that category. But you’ve probably seen these. The most common example is “no vs. known,” as in “There is no evidence of cancer,” or “There is known evidence of cancer.” Scary, right? Here’s another example of how the misinterpretation of a single letter can cause a report to mean the exact opposite of what is intended: “There is now evidence of cancer,” or “There is not evidence of cancer.”


Whether you know it or not, your documentation is likely full of mistakes like these. How do I know that about your reports? Well, if you’re reading this, you’re probably concerned about documentation snafus; that concern led you to SpeechCheck. And, as a broad generalization, we find mistakes like these in the documentation for every hospital with which we work. More simply: If your documentation is 100% accurate all the time, you wouldn’t be taking the time to read this post.


At SpeechCheck, we believe that all patients deserve imaging reports that are error-free. Our baseline assessments give a very accurate picture of our clients’ documentation quality, and our training helps radiologists reduce errors almost immediately. Imagine the peace of mind this could bring! Better yet, think about the results of a documentation quality improvement initiative. Wouldn’t it be great to know that the quality of your reports was improving over time?


Authenticated radiology reports are the end work product of one of the most expensive processes in healthcare. Can one letter really be that important to patient care? Is documentation quality really that big of a deal? At SpeechCheck, we think sew.

Posted by: Ken Schafer

Back to Top