APIII - Advancing Practice, Instruction & Innovation Through Informatics

A Simple Easy-to-managed-by-Registrar Cancer Registry
Expert System from Pathology Report

Polun Chang, PhD1 (polun@ym.edu.tw), Ya-Fen Liang, MSN, CTR, RN2, 1 Institute of Health Informatics and Decision Making, National Yang-Ming University, Taipei, Taiwan/ROC;  2 Department of Tumor Center, Changhua Christian Hospital, Changhua, Taiwan/ROC

Context:  The techniques used to develop the cancer registry automation applications often are not easy to be understood and maintained by the cancer registrars, who are usually less competent in computer literacy.  The core component of these intelligent applications is the knowledge base to parse the semi-structured pathology reports.  Since the cancer registrars are the best experts to define this knowledge base, a system with easy, simple and familiar interface to maintain this base should make the system more useful.

Technology:  Excel spreadsheet was chosen to design the system because it has been most familiar to and widely used by healthcare workers.  It can support simple but good enough database functions and is powerful in analysis.  More importantly, it is embedded with the VBA programming tool which can be used to code the parsing algorithms and to exchange files with the HIS (Hospital Information Systems).
 
Design:  One experienced certified tumor registrar, who is a nursing informatics MS and in charge of a 12-member tumor center of 1500-bed medical center, and one health informatics professional, who is familiar with Excel VBA programming, were teamed up to use an intensive prototyping approach to develop the system.  517 free-text pathology reports of three hundred and five breast cancer patients in 2005, were retrieved from the HIS system for content analysis to design the parsing algorithms.   Another 160 reports from one hundred and two breast cancer patients in 2006 were used to test the system. 
 
Results:  The system is composed of one parsing and two file exchange modules.  There are six Excel worksheets: (1) the Report to store reports for parsing, (2) the Key Word/Phase to include 126 key words and phases and how far the words nearby to search, (3) the Exception Checking to examine the format of the reports and to identify the cases with different formats or cannot fully parsed, (4) the In-Processing to store the in-processing results and the list of 45 final parsed words or phases, and (5) the Auto-Registry to lay up the rules for analyzing the final words and phases to determine the final coding and (6) the Result to store the final 16 coding items.  98% of reports were correctly coded with high registrars’ satisfaction. 
 
Conclusion:  The results encourage us the feasibility and effectiveness of designing a cancer registry automation system manageable by the cancer registrars. 

Keywords: Expert System, Cancer Registry, Pathology Report, Automation, Excel, VBA, User Oriented