September 16, 2019 | 04.30 PM EDT
Documentation of clinical encounters is almost as old as documented civilization itself. While it was primarily used as a means of teaching and research in the olden days, it has been attributed to a lot of regulatory significance in the modern age. The process of transcription itself has undergone a tremendous change from how it was written on papyrus by hand a few millennia ago, to how it is almost entirely automated in the present day.
In fact, medical transcription has always embraced modernization and has been at the forefront of incorporating new technologies within the process. Voice recording found very early application in medical transcription as it meant that the providers no longer needed to document their patient’s information themselves, but could get the support of transcriptionsists to transcribe their recordings. However, the advent of the Internet helped providers to share their dictation notes to remote transcriptionists from across the globe. They could get their work done irrespective of the time and distance between the providers and transcriptionists.
Later, speech-to-text conversion, one of the booming technological advancements of the latter half of the 20th century came along. This technology again found immediate application in the field of medical transcription, as it made the lives of transcriptionists a little easy. However, the next evolution occurred when the internet allowed voice recordings to be streamed over to a remote transcriptionist. It helped the healthcare industry to access highly trained medical transcriptionists from the global workforce.
In the modern healthcare setup, physicians are expected to update the patient’s Electronic Medical Record (EMR) after each visit, entering information by hand into multiple fields on the form, which can sometimes stretch to pages. Although the EMRs are electronic documents, the process of manually filling out the patient’s details is tedious, time consuming and unproductive for the provider. The Forbes reported that the providers spent nearly 50% of their work time on EHRs based on a research study by National Institutues of Health.
Clinical documentation before the advent of Artificial Intelligence was mundane and time-consuming. However, machines were ‘taught’ to listen to a voice, process the voice input and convert the speech to text for the transcriptionists. Though these speech to text conversions were not always accurate, the system continued to learn from the edits performed by the human transcriptionists. Feedback loop and steady accuracy improvement, trained in part by humans, has been in practice at least for the past 15 years.
On the other hand, Natural Language Processing (NLP) algorithms are used to extract meaningful information from all the raw unstructured data that is available in a typical medical transcript. Software is ‘trained’ to read a sentence on the transcript and infer an actionable piece of knowledge about the patient from it, which then populates into the corresponding field in the patient’s Electronic Medical Record (EMR) automatically. NLP is built on a stack of multiple technologies and also requires an extensive ontology library. Consequently, it’s important to employ highly skilled data scientists in order to streamline the NLP workflow and take advantage of its diverse capabilities.
CascadeMD leverages its robust AI engine to extract meaningful information from unstructured transcription text, with autopopulation of EHRs as one of the several possibilities. Automation is clearly the path ahead for clinical documentation and AI and NLP are well on their way to making this path a successful and sustainable one.