THE DEVELOPMENT OF SYNOPTIC REPORTS FROM FREE TEXT CONTENT OF ARCHIVAL PATHOLOGY REPORTS GENERATED IN THE ANATOMIC PATHOLOGY LABORATORY INFORMATION SYSTEM
Purpose of Synoptic Reports Structure the information Structure the information Structured form facilitates research queries Structured form facilitates research queries Structured form facilitates better patient care Structured form facilitates better patient care Structured form facilitates better sharing of information between institutions Structured form facilitates better sharing of information between institutions
Example Synoptic Page
Conversion of Archived Systems An institution that changes from free formatted pathology reports to synoptic reporting may wish to convert all archived reports to synoptic form An institution that changes from free formatted pathology reports to synoptic reporting may wish to convert all archived reports to synoptic form Is it better to convert archived data by hand or automatically with respect to: Is it better to convert archived data by hand or automatically with respect to: AccuracyAccuracy TimeTime
Computer Methods Search for key words Search for key words Use auto-coding vocabularies so that all (or most) variations of linguistically equivalent phrases are placed in a uniform phrase for proper pattern matching in searches Use auto-coding vocabularies so that all (or most) variations of linguistically equivalent phrases are placed in a uniform phrase for proper pattern matching in searches Use negation algorithms Use negation algorithms
Special Challenges (What is the Tumor Site for this case?) Part A. RECTOSIGMOID COLON ULCER, BIOPSY: --INFILTRATING MODERATELY DIFFERENTIATED ADENOCARCINOMA, CONSISTENT WITH COLONIC ORIGIN, SEE NOTE. Part C. COLON, LEFT, HEMICOLECTOMY: --INFILTRATING MODERATELY DIFFERENTIATED ADENOCARCINOMA OF RECTOSIGMOID COLON, SEE NOTE.
Which is “Whipple” and which is “Whipple Pylorous Sparing”? Received fresh for intraoperative consultation, labeled with the patient's name, number and "Whipple," is a bloc section of organs from a Whipple procedure. The specimen consists of a 24.0 cm. in length duodenum and a 6.2 x 4.3 x 4.0 cm. remnant of attached pancreas, and a portion of distal stomach, 6.5 x 2.5 cm. Received fresh for intraoperative consultation, labeled with the patient's name, number and "Whipple," is a bloc section of organs from a Whipple procedure. The specimen consists of a 24.0 cm. in length duodenum and a 6.2 x 4.3 x 4.0 cm. remnant of attached pancreas, and a portion of distal stomach, 6.5 x 2.5 cm. Received fresh for intraoperative consultation, post-fixed in formalin, labeled with the patient's name and number, is an en bloc resection of organs from a Whipple procedure. The specimen consists of a 21.5 cm. length of duodenum, a 1.5 cm. length of distal common bile duct, and an 8.8 x 6.8 x 6.7 cm. remnant of attached pancreas. Received fresh for intraoperative consultation, post-fixed in formalin, labeled with the patient's name and number, is an en bloc resection of organs from a Whipple procedure. The specimen consists of a 21.5 cm. length of duodenum, a 1.5 cm. length of distal common bile duct, and an 8.8 x 6.8 x 6.7 cm. remnant of attached pancreas.
Design of Experiment To determine the feasibility of automated conversion of archived reports, we collected all records for the year 2005 and converted them to synoptic form, both manually and by computer text processing. We then compared the results for accuracy and the time to process. The organ sites we studied were: Pancreas Pancreas Kidney Kidney Colon Colon Prostate Prostate Breast Breast
Estimated Accuracy for Pancreas Specimen type Tumor Site Tumor size Histologic type Histologic Grade 23/2592%18/2572%19/2576%24/2596%24/2596%
Estimated Accuracy for Kidney Laterality Tumor Site Tumor Size Histologic Type 27/27100%21/2778%7/2726%24/2789%
Estimated Accuracy for Colon Tumor Site Tumor Size Histologic Type 40/7553%42/7556%74/7599%
Estimated Accuracy for Prostate Histologic Type Gleason Primary Gleason Secondary Gleason Total 49/5098%48/5096%48/5096%48/5096%
Estimated Accuracy for Breast Bloom- Richardson grade Mitotic Score Angiolymphatic Invasion 100%100%100%
Estimated Time for Computer Conversion The average time to develop the text processing code was approximately 4-6 hours per organ. The first organ took the longest time, the others were re- written from the first.
Estimated Time for a Trained Pathologist to convert archive reports to synoptic reports Total Time Number of Cases Average Time PancreasN/AN/AN/A Colon9:27:48727:53 Breast4:46:46485:58 Kidney3:57:23327:25 Prostate 3:38:30 3:38:30307:17
Conclusion Less than cases, a trained pathologist is quicker. Less than cases, a trained pathologist is quicker. More than cases, automated text processing is quicker. More than cases, automated text processing is quicker. Attempts to improve the accuracy will increase this time. Attempts to improve the accuracy will increase this time. For archived systems with a very long history and a large volume of patients, automated processing may be preferred. For archived systems with a very long history and a large volume of patients, automated processing may be preferred.
Acknowledgements We wish to thank the following for their participation in this project: Ashok Patel Ashok Patel Doug Hartman Doug Hartman John Gilbertson John Gilbertson Robert Lanese Robert Lanese Rajnish Gupta Rajnish Gupta