Text analytics can save lives

I’m just back from the Big Data Summit in New York where I was hoping to hear folks talking about unstructured data. After all, it makes up 80 percent of all data, according to IBM.

I was not disappointed. Craig Rhinehart of IBM talked about finding treatment insights in unstructured data, which he illustrated with two case studies. Seton Health Care in Texas wanted to know why a significant percentage of heart patients were being re-admitted to the hospital. IBM found the answer by looking at patient records and doctors’ notes. But first they polled the doctors to see what they thought caused the returns. The No. 1 reason wasn’t on the list. It was, however, in the notes of the patient records. It was a small value in a long list of small values on a sheet that was noted by hand. Clerks often missed the item when inputting the information into patient electronic records. So, again, the primary indicator of patient re-admission was tracked in text notes and discovered only after IBM surfaced the information through natural language processing text analytics.

Rhinehart second case study was about the UNC Healthcare where some patients with abnormal cancer screening results were not getting follow up care. Rhinehart was careful to say that this happens every day in every healthcare system, so UNC isn’t careless. Instead, the people who work there are human.

It seems that doctors were putting follow up indicators in patient notes, and clerks who were supposed to update the field were missing it when inputting data. Downstream follow up actions were triggered by the structured field that needed to be populated by the clerks. IBM used natural language processing to extract the information about the abnormal screenings and automatically send alerts for doctors to proactively follow up with patients. Read Rhinehart’s post on the issue here.

Better, more complete insights can be gained through looking at unstructured data. And that data can help save lives.