Back when I was an undergraduate biochemistry major, I used to fantasize about blasting down into my own cells to watch the molecular reactions I was studying in class. Imagine: How cool would it be to see every step in the conversion of a single glucose molecule into energy that powers your muscles or brain? And, I figured, if I could watch that, I could also watch all the other as-yet-undiscovered metabolic processes in my cells and understand them perfectly!
I know - a really nerdy fantasy.
But some physicians, as it turns out, harbor a similar fantasy: They wish to learn from every single patient they see. Like my cells, patients' medical histories are a potential treasure trove of information - if only there was a good way to see it. Not sure what X treatment would do in Y clinical situation? Pull up records for the last 100 or 1,000 patients who went through the problem and run a quick statistical analysis to see how they reacted to different types of care. It wouldn't be as exacting as a randomized clinical trial, but many kinds of patients (those with rare diseases, for instance) will never be studied in such trials. If only …
Now, as a New England Journal of Medicine perspective piece from a team at Lucile Packard Children's Hospital describes, this fantasy is beginning to come true. A team at the hospital recently did a quick, efficient analysis of de-identified data from about 100 prior patients with a rare pediatric condition, allowing them to make a difficult clinical decision about a gravely ill child. The team says it marks the first time electronic medical records have been harnessed for this type of real-time decision making.
As the press release I wrote about the case describes:
The patient in this case was a 13-year-old girl admitted to the hospital with severe inflammation of the kidneys and pancreas that had developed as a complication of lupus. Her condition is rare in children, and no clinical trials or other medical literature addressed the best course of care for her unique situation. The medical team, led by rheumatologist Jennifer Frankovich, MD, realized the girl was at risk for blood clots that could cause organ failure or death. However, the anticoagulant drugs that would prevent blood clots would also increase the risk of bleeding from invasive medical procedures, such as the kidney biopsy the girl needed.
Frankovich, who is also an instructor of pediatric rheumatology at the Stanford University School of Medicine, consulted her colleagues about what to do. They said anticoagulation probably wasn’t needed, but had little evidence to back up their advice. That worried Frankovich, who couldn’t shake her anxiety about the patient.
“She was sick, and I felt like in the next three days we could make her a lot sicker or a lot better,” said Frankovich, the first author of the paper.
Frankovich's analysis showed that it was probably best to give the anticoagulant drug. She did, and the patient recovered and is now doing well.
The analysis was possible because of Stanford's unusually capable electronic medical record; at most hospitals, the EMR is still used just like a paper chart, with no ability to learn from aggregate patient data. But that's likely to change:
The ability to learn from electronic records is getting faster, the team said. Thanks to advances in the Packard Children’s electronic medical record and Stanford’s accompanying research tools, the search that took Frankovich four hours when this patient was seen in early 2010 would take about an hour today.
Frankovich added, “I expect in the future that you’ll be able to look at aggregate patient data during rounds, so you can make more-informed patient care decisions right there.”
So amazing. Now about my biochemistry fantasy …
