Over the past couple of months I’ve taken the time to review my electronic records from all of my major providers. Unfortunately, I’ve not been very impressed.
The vast majority of the data I have received is filled with fluff. In some cases this has been due to “cloning” – the practice of taking a previous encounter and copying it wholesale into the current visit. More worrying though is the prevalence of nonessential data in the narrative. I’m starting to wonder how any medical professional could actually work with these documents.
Unfortunately, it’s not just me. My job is the Chief Technology Officer at eHealth Technologies – a major health information clearinghouse for continuity of care. Each month, we go out to over 20,000 locations in the United States and collect patient records in both discrete and unstructured formats. We see a tremendous amount of data, and in my discussions with staff who work in the HIPAA restricted zones, I’ve heard of many similar issues.
How do we deal with this problem? In short, we use the process of aggregating data into a physician friendly format that we call the SmartPDF. Let me give you a very quick overview.
The first thing to understand is that in any case there is a tremendous amount of duplication. It is not uncommon to see the same documents repeated over and over again in the patient’s chart. This has the potential to be a tremendous time waster for a physician who has to look through all the pages or face lawsuits. So, we use a semiautomated approach to get rid of these.
Second, not all data is created equal. The physician typically wants to see the most recent (and often most relevant) data first, grouped by the document type. Being able to jump around easily within the data is critical and we provide navigation features that greatly simplify this process.
Third, some clinical terms are particularly relevant. Our medical director has carefully selected a dictionary of clinical phraseology for each disease type that our customers treat, and through the use of OCR and natural language processing we are able to highlight these terms, present them in the clinical summary, and make them navigable.
There is a lot more to data aggregation from a technology perspective, but these three core processes allow us to extract the most relevant data out of the massive compendium of documents that are involved in, for example, an oncology, or transplant case.
I’m doing my part to get rid of fluff in medical records. Are you?