In the USA already included billions in projects to introduce electronic
medical records. The quality of care has thus risen but not as expected.
One problem: The data can be automatically analyzed to date are
difficult to problems in the health system to get on the track.
A new study shows that methods could help in the field of natural
language speech recognition, known as computational linguistics, in the
processing of patient information, reports [2] Technology Review in its
online edition. It took the U.S. researchers in those areas previously
considered particularly complex: the part of the medical record where
physicians leave their medical notes.
On examination, the medical faculty of the Harvey Murff [3] at
Vanderbilt University has been carried out, efforts were made to
material that surgeons earned about possible complications after
operations on the files. To feed its database, the researchers analyzed
nearly 3,000 anonymised medical records of patients who underwent
surgery in six hospitals in the U.S. Army.
It was about complications such as pneumonia, sepsis, venous thrombosis,
pulmonary embolism and heart attack. Increased problems after surgery
can be a sign that an institution does not monitor sharp enough basic
safety rules. However, such problems are now often only when recheck by
hand. "We wanted to mimic what would a human expert and then scale up
this process - and as efficiently as possible," said Murff.
Your finished system is amazingly accurate: The software was seen
between 80 and 90 percent of all complications in the records as good as
a man. Other automated methods, based for example on the reading of
billing codes, were clearly inaccurate. While Murff system recognized
for example, up to 82 percent of all cases of kidney failure, it worked
on billing codes in only 38 percent of cases.
No comments:
Post a Comment