NEW YORK (Reuters Health) – Free-text searches of electronic medical records (EMRs) perform better than administrative data codes in identifying postoperative complications, a study in the August 24/31 issue of the Journal of the American Medical Association (JAMA) indicates.

This finding is not all that surprising, Dr. Harvey J. Murff, of the Veterans Affairs Medical Center and Vanderbilt University, Nashville, Tennessee, told Reuters Health. “Administrative codes are predominately used for reimbursement purposes; they were not originally developed for use as an indicator of medical care quality,” he explained in an email.

Dr. Murff and colleagues evaluated a natural language processing-based approach to identify surgical complications in the EMRs of 2,974 patients who had surgery at Veterans Health Administration medical centers from 1999 to 2006.

They focused on six postoperative complications: acute renal failure requiring dialysis, deep vein thrombosis, pulmonary embolism, sepsis, pneumonia, and myocardial infarction. These complications were identified through medical record review by trained nurses as part of the VA Surgical Quality Improvement Program.

For the data set, the percentage of postoperative acute renal failure requiring dialysis was 2%; for pulmonary embolism, 0.7%; for deep vein thrombosis, 1%; for sepsis, 7%; for pneumonia, 16%; and for heart attack, 2%.

The researchers calculated the sensitivity and specificity of the free-text search approach to spot these complications and compared its performance with patient safety indicators that use administrative discharge codes.

In general, the natural language processing-based approach had higher sensitivities and lower specificities than did the administrative data codes, the study team found.

“The increase in sensitivity of the natural language processing-based approach compared with the patient safety indicator was more than 2-fold for acute renal failure and sepsis and over 12-fold for pneumonia,” they note. Specificities were 4% to 7% higher with administrative codes than the natural language processing approach.

Free-text searches correctly identified 82% of acute renal failure cases compared with 38% for discharge codes. The corresponding percentages were 59% vs. 46% for vein thrombosis, 64% vs. 5% for pneumonia, 89% vs. 34% for sepsis, and 91% vs. 89% for MI.

“Both natural language processing and patient safety indicators were highly specific for these diagnoses,” the investigators say.

The author of a linked commentary calls the study “innovative” by going beyond the traditional uses of the EMR (i.e., helping clinicians deliver guideline-based care and reduce medication errors) to show that natural language processing, when applied to electronic data, can help clinicians track adverse events after surgery.

“To many readers, the topic may appear esoteric, but its significance should not be underestimated,” writes Dr. Ashish K. Jha from the Harvard School of Public Health in Boston. Instead, these findings suggest that EMRs “can transform health care delivery.”

Dr. Jha also notes that the authors tested dozens of permutations and combinations of syntax to identify the best strategy for finding these six complications in the EMR.

“We knew that the information we needed was contained within the medical record; it was just a matter of getting it out,” Dr. Murff commented. “It is an iterative process,” he said, “and takes some trial and error while fine tuning the search algorithms. Indeed, the ones we report on still could use some improvements, but as more and more institutions adopt EMRs this approach could be a useful tool for quality improvement,” he added.

For now, the major barrier to the more widespread use of free-text searchers is the current limited use of EMRs in healthcare, Dr. Murff told Reuters Health.

“Many healthcare systems still rely on paper records and would not be able to benefit from these ‘electronic’ chart reviews. We hope the results of our study help call attention to another potential benefit of electronic medical records and only further encourage their adoption,” Dr. Murff said.

Dr. Jha makes the point that while the promise of natural language process is “substantial, its benefits will not be realized without considerable new investment in research and development.”

“Although there are private-sector companies capitalizing on the benefits of natural language process to help clinicians and organizations improve care delivery, the federal government can play a helpful role by funding the basic research needed to launch this field forward,” Dr. Jha concludes.

JAMA 2011.