Searching electronic medical records using natural language processing identified postoperative complications better than more commonly used data codes, researchers say.
Led by Dr Harvey Murff, of the Veterans Affairs Medical Center and Vanderbilt University, researchers conducted a study of database searching to identify a set of 20 measures of patient safety indicators based on discharge coding to screen for potential adverse events that occur during hospitalization. The study appears in today’s Journal of the American Medical Association.
"Using natural language processing with an electronic medical record greatly improves postoperative complication identification compared with the patient safety indicators, and administrative-code based algorithm,” the researchers wrote.
The study compared natural language processing with current methods based on administrative data codes to screen the medical records of 2,974 patients who underwent inpatient surgical procedures at six Veterans Health Administration medical centers between 1999 and 2006.
Natural language processing (NLP) is a field of computer science and linguistics concerned with the interactions between computers and human (natural) languages. Database queries using NLP allows broader, more flexible searching that can process daily progress notes, microbiology reports, or imaging reports in addition to structured fields. Administrative code queries only search structured field data in an electronic health record.
In the study the researchers compared the ability of the two query strategies to identify postoperative occurrences of acute renal failure requiring dialysis, deep vein thrombosis, pulmonary embolism, sepsis, pneumonia, or heart attack as part of the VA Surgical Quality Improvement Program. All six hospitals used the same electronic medical record system.
Overall of 1,924 patient records queried, 39 patient (2 percent) experienced acute renal failure requiring dialysis, 0.7 percent (18 of 2,327) had pulmonary embolism, 1 percent (29 of 2,327) experienced deep vein thrombosis, 7 percent (61 of 866) had sepsis, 16 percent (222 of 1,405) had pneumonia, and 2 percent (35 of 1,822) experienced myocardial infarction.
Natural language processing correctly identified 82 percent of acute renal failure cases compared with 38 percent for patient safety indicators queries. Similarly NLP identified 59 percent of the venous thromboembolism compared to 46 percent for safety indicators. For pneumonia NLP identified 64 percent vs. 5 percent, 89 percent of the sepsis cases vs. 34 percent for safety indicators. NLP correctly identified 91 percent of the postoperative myocardial infarction compared to 89 percent for conventional query. Both natural language processing and patient safety indicators were highly specific for these diagnoses.
The authors suggest that a natural language processing-based approach offers several potential advantages over administrative-code based strategies to identify healthcare quality concerns.
"First is the flexibility of the approach to meet the individual institutional needs. Once documents have been processed, different approaches and query strategies to identify a specific outcome can be implemented at a relatively low programming effort using standard database query applications,” they wrote. “Second, as opposed to administrative codes, search strategies using daily progress notes, microbiology reports, or imaging reports could be monitored on a prospective basis. Thus, this approach could potentially identify complications while a patient is still in the hospital, which could greatly facilitate real-time quality assurance processes.”
They added that in systems with highly integrated EMRs, prospective surveillance could be extended to the outpatient setting for individuals remaining with the healthcare system.
Writing in an editorial accompanying the research study, Dr. Ashish Jha of the Harvard School of Public health praised the study for showing the promise of natural language processing to provide new insight into quality improvement efforts, but said it needs to be extended to disciplines other than surgery if the benefits are to be realized and called on government funding to accelerate that development.
“Although there are private-sector companies capitalizing on the benefits of natural language process to help clinicians and organizations improve care delivery,” he wrote, “the federal government can play a helpful role by funding the basic research needed lo launch this field forward."