Natural Language Processing Programs Can Effectively Measure Colonoscopy Quality According to University of Pittsburgh Researchers
PITTSBURGH, May 24, 2012 – Software that analyzes written text, known as natural language processing (NLP) programs, can effectively measure colonoscopy quality in an inexpensive, automated and efficient manner, according to researchers at the University of Pittsburgh School of Medicine, whose findings appear in the June edition of Gastrointestinal Endoscopy.
“Gastroenterology specialty societies have called for physicians to regularly measure the quality of colonoscopy screening. However, physicians face challenges doing this because it simply takes too much time to review each medical record,” said lead author Ateev Mehrotra, M.D., assistant professor in the Division of General Internal Medicine at the University of Pittsburgh School of Medicine and a researcher at RAND, a nonprofit research organization. “Our NLP computer program can replace a human and quickly analyze electronic medical records, making the process simple and cost-effective while improving the quality of care.”
In this study, more than 24,000 colonoscopy reports from a two-year period contained in the digital health records at UPMC were analyzed. Researchers used NLP to mine the data contained in the electronic medical records to measure and report on the quality of care for seven measures. These included an indication of the patient’s physical status, documented informed consent of patient, description of the quality of bowel preparation, notation of bowel landmarks, detection of adenoma, notation of the time the scope was withdrawn, if indication for colonoscopy is chronic diarrhea, and whether or not a biopsy sample was taken.
The analysis of colonoscopy reports revealed that the documentation of most quality measures was poor with a wide range of performance across hospitals and physicians—including a threefold variation in the adenoma detection rate among doctors. In addition, physicians who used a structured program to document their colonoscopy had a higher quality score than those who used dictated reports.
“Our study indicates that NLP could be a means of making quality reporting on colonoscopies more common and lay the foundation for efforts to improve care,” said Robert E. Schoen, M.D., M.P.H., study co-author and professor in the Division of Gastroenterology and Hepatology at the University of Pittsburgh School of Medicine.
Researchers were supported with funding from the National Institutes of Health, and a pilot grant from the University of Pittsburgh Clinical and Translational Science Institute and the RAND-University of Pittsburgh Health Institute.
Co-authors of the study are Melissa Saul, M.S., and Henk Harkema, Ph.D., from the University of Pittsburgh; Evan S. Dellon, M.D., from the University of North Carolina School of Medicine; and Faraz Bishehsari, M.D., from Northwestern University Feinberg School of Medicine.