Po-Hsun Cheng
†*, Jin-Shin Lai*, Fong-Ming Shyu
‡*, Heng-Shuen Chen**, Jer-Junn Luh
§, Sao-Jie Chen
†‡† Graduate Institute of Electronics Engineering, ‡ Department of Electrical Engineering, National Taiwan University; Departments of * Information Systems, ** Family Medicine and Medical Informatics, and
§Physical Medicine and Rehabilitation, National Taiwan University Hospital and College of Medicine, Taipei, Taiwan.
H
TIJ
Correspondence and reprint requests: Sao-Jie Chen, PhD, Graduate Institute of Electronics Engineering, National Taiwan University, No. 1, Sec. 4, Roosevelt Road, Taipei, Taiwan, 106.
E-mail:[email protected].
© The Journal on Information Technology in Healthcare 2003; 1(5): 331–342
ABSTRACT
Objective: This paper discusses the necessity and decision to implement a Health Level Seven (HL7) standard to enable healthcare data exchange between hospital information sys-tems (HIS), laboratory information syssys-tems (LIS), and automatic medical instruments (MIs).
Design: The major design challenge in developing intelligent laboratory information serv-ices is to enable distributed and heterogeneous medical instruments to be quickly, easily and cheaply connected to the LIS and the HIS.
Setting: University Hospital in Taiwan.
Methods: We followed the HL7 v.2.4 standard to define the related laboratory HL7 mes-sages and to use the HL7 Interface Engine (IE) to communicate between heterogeneous systems and MIs. We used these to connect three Toshiba biochemistry machines to the LIS and HIS.
Results: Connections were accomplished in less than one week compared to a previous connection time of at least four weeks. The new system is able to process messages at more than twice the speed of the old system. It is currently running smoothly, handling almost 10,000 tests each day. Data exchange between the MIs, LIS and HIS is reliable, accurate and complete.
Conclusions: We have demonstrated that our approach is feasible and works in practice.
We now plan to extend this approach to other medical instruments. Once we are assured of its reliability and applicability to a wide range of different MIs, we will promote our ap-proach to other medical institutions. A flow chart is presented to aid other healthcare insti-tutions decide if they should adopt HL7 to connect their medical instruments to their laboratory and hospital information systems.
Cheng, Lai, Shyu et al
332 The Journal on Information Technology in Healthcare 2003; 1(5): 331–342
INTRODUCTION
Hospital software development is becoming increasingly complicated. Hospital system requirements for web-based, client/server and legacy applications, com-bined with support for multiple platforms, and sophisticated end-user functionalities have forced information system developers to pursue and adopt new approaches including integrated architectures, such as medical information exchange and medical application services1,2
One of the major problems encountered with data exchange between different healthcare information systems is the lack of a uniformly accepted standard.
Unlike standards used for general data exchange in businesses such as insurance and banking, these standards have to be directly designed for healthcare data exchange. As a consequence there are several popular data exchange standards in use around the world. Some of these and their developers are listed below:
Arden Syntax from Columbia University and HL73
Clinical Context Object Working Group (CCOW) from HL74 Clinical Document Architecture (CDA) from HL75,6
ASTM-E31 from the American Society for Testing and Material (ASTM)7 DICOM-3 from the National Electrical Manufacturers Association (NEMA)8 SNOMED from the College of American Pathologists (CAP)9
LOINC from the Regenstrief Institute10
CORBAmed from the Object Management Group (OMG)11 ActiveX for healthcare from Microsoft12
X12N from the Accredited Standards Committee (ASC)13
The choice of which standard to adopt requires careful consideration. Use of a well-defined and widely-acceptable medical data exchange standard is crucial to ensure both long-term use and to allow for the possibility of future develop-ment14,15. HL7 has been specifically developed for transmitting text messages among healthcare information systems and is widely used around the world. This standard was thoroughly tested at the National Taiwan University Hospital (NTUH) before the decision to incorporate it as the standard for the hospital’s information infrastructure was made16.
Careful consideration also needs to be given to the choice of interface used for connecting the laboratory information system to the heterogeneous medical in-struments (MIs). We initially proposed using our own NTUH Laboratory Instrument Communication Interface (NTUH/LICI) standard for this purpose.
This standard was developed over a 3-year period and implemented at NTUH in June 1999. As shown in Figure 1, it uses two file queues to handle the data exchange between the Laboratory Information System (LIS) and MIs. At NTUH it has been used to connect 40 MIs with our LIS. The process rate is about one message per second. The size of each command sent from a PC to an MI is approximately 130 bytes and the size of the result sent from the MI to the PC is
Laboratory Data Exchange Using HL7-Interface Engine
The Journal on Information Technology in Healthcare 2003; 1(5): 331–342 333 approximately 415 bytes. This file queue process mode is suitable for slow-speed MIs, but not for high-speed MI requirements. Our experience has also shown that each time a new MI is connected it takes at least one man month for planning, coding and testing. After careful consideration we decided that it would be better to use an internationally accepted interface. We felt that in the long-term this would offer significant advantages including simplifying and speeding up the connection process.
METHODS
In January 2003 our largest automatic biochemistry instrument, the Hitachi 7450, was replaced by three new Toshiba models (two TBA-200FR and one TBA-120FR).
Using the accepted HL7 standard we managed to connect these three MIs to the LIS and HIS within one week. The connection pathways are shown in Figure 2, and technical details are given in Appendix A.
The steps involved in processing a sample can be summarised by the following four stages:
1. A doctor inputs laboratory orders into the HIS, and the HIS passes the message to the LIS.
2. When the specimen is received, a laboratory technician logs in the labora-tory order entry in the HIS.
3. LIS transmits the information for a specific order to a specific MI which processes the specimen, generates a report and sends it back to the LIS.
Figure 1. Laboratory Instrument Communication Interface (LICI) at National Tai-wan University Hospital (NTUH)
This uses two file queues to handle the data exchange between the Laboratory Information System (LIS) and the Medical Instrument.
Front-end
Cheng, Lai, Shyu et al
334 The Journal on Information Technology in Healthcare 2003; 1(5): 331–342
4. The laboratory technician or doctor releases the laboratory report to the HIS.
Data exchange using the HL7-Interface Engine (HL7-IE) gateway solution between the HIS, LIS and Toshiba MIs was evaluated for 13 days in February and March 2003. The results are shown in Figures 3 and 4.
RESULTS
All three Toshiba MIs were connected to the LIS and HIS without difficulty within a one week period in February 2003. Since this connection was the first prototype attempt in Taiwan, the Toshiba MI connection interface implementations were provided free of charge by the vendors. However, due to the quicker connection time the new HL7 connection method should be cheaper than the old connection method which costs approximately US$2,000 for each MI. Currently the system is working well handling an average of 31,267 tests per day for all laboratory orders and 9,725 tests per day for the Toshiba MIs. Data exchange between the MIs, LIS Figure 2. Detailed data exchange flow between the Hospital Information System (HIS), Laboratory Information System (LIS) and Medical Instruments (MIs)
The dashed vertical line delineates the use of two different solutions. The right-hand side uses the HL7-IE (Health Level 7-Interface Engine) gateway solution and the left-hand side uses the Sybase gateway solution. IBM-HPx: represents the use of HL7 programs at IBM (International Business Machines) mainframes; IBM-APx: represents the use of adapter programs at IBM mainframes; LIS-SPx: stored procedures at LIS; LIS-APx: adapter programs at LIS; LIS-HPx: the HL7 programs at LIS; and TSB-HPx: the HL7 programs for the Toshiba medical instrument system. Appendix B gives a table comparing the HL7-IE gateway and the Sybase gateway solutions.
ORU^R01 Send lab order & print sheets after barcode
Lab Temporary store at PostDetailtable, then forward to LabDetail table LI S-SP3
Laboratory Data Exchange Using HL7-Interface Engine
The Journal on Information Technology in Healthcare 2003; 1(5): 331–342 335 and HIS is reliable, accurate and complete.
Figure 3 shows the number and size of HL7 laboratory report messages passing between the LIS and Toshiba MIs on 13 individual days. The LIS MessageOut represents data passing from the Toshiba MIs to the LIS whereas the LIS MessageIn represents the corresponding acknowledgement message from the LIS to the Toshiba MIs. In the absence of errors or reject messages, the number of messages between these two systems should be exactly the same. The number of messages sent each day between these two systems is identical except for one day (day 10) where they differ by 1 message. There is, however, a notable difference in the size of the mes-sages. Messages sent from the MIs to the LIS are approximately 2.7 times bigger than messages sent from the LIS to the MIs (337 bytes vs. 125 bytes respectively).
Figure 4 shows the number and size of HL7 laboratory order messages passing between the HIS and the LIS over the same 13 days. The HIS MessageIn repre-sents laboratory order messages passing from the HIS to the LIS and the HIS MessageOut represents the corresponding acknowledgement message from the LIS to the HIS. There are small differences in the number of messages passing between these two systems on three days (days 5, 10 and 11). There is a much larger discrepancy of 23 messages on both days 7 and 8. However, it should be noted that there were more messages in than out on day 7, whereas on day 8 the converse is true. It is possible that this discrepancy is an artefact related to the timing of collection of data as the number of messages passing between these two systems is exactly the same for the two days combined. Each message passing from Figure 3. Data exchange using HL7-IE gateway for laboratory tests between the LIS and MIs
0 5 0 0 0 1 0 0 0 0 1 5 0 0 0
8 9 4 8 9 2 3 0 8 6 7 3 8 1 7 2 1 3 7 1 7 6 3 1 8 8 0 3 8 6 5 6 1 3 4 5 9 9 0 6 1 0 1 5 9 8 4 8 9 2 4 3 8 9 4 8 9 2 3 0 8 6 7 3 8 1 7 2 1 3 7 1 7 6 3 1 8 8 0 3 8 6 5 6 1 3 4 5 9 9 0 7 1 0 1 5 9 8 4 8 9 2 4 3 1 0 9 2 1 1 2 7 1 0 5 9 9 9 8 1 6 7 4 9 3 2 8 6 9 1 2 6 2 1 6 4 3 1 2 0 9 1 2 3 9 1 2 0 2 1 1 2 8 2 9 5 3 3 0 3 6 2 8 5 9 2 6 9 4 4 5 1 6 2 5 1 3 2 3 8 1 3 3 6 9 4 4 3 6 3 2 6 4 3 3 4 5 3 2 4 4 3 0 4 5
1 2 3 4 5 6 7 8 9 1 0 1 1 1 2 1 3
Log Date
Message Quantity (No. & Kbytes).
LIS MessageIn (No.) LIS MessageOut (No.) LIS MessageIn (KB) LIS MessageOut (KB)
Cheng, Lai, Shyu et al
336 The Journal on Information Technology in Healthcare 2003; 1(5): 331–342
the HIS to the LIS is roughly 3.5 times the size of the message passing from the LIS to the HIS (394 bytes vs. 114 bytes respectively).
DISCUSSION
Automation of healthcare systems and processes within a hospital is desirable to improve the efficiency of healthcare. This is a major challenge due to a combina-tion of disparate distributed systems, a wide variety of data both in text and image format and no universally accepted standards for data exchange. A fundamental first step in dealing with the problem is realising and dealing with the challenges of system selection, implementation and maintenance.
As mentioned earlier there are a number of different healthcare data transfer standards in use around the world. To connect a wide range of medical instru-ments it is necessary to choose standards that will allow data exchange in text format (e.g. biochemistry results), image format (e.g. x-rays) or a combination (e.g. electrocardiogram). In the healthcare setting, well recognised and widely used standards are HL7 for text transfer, and Digital Imaging and Communica-tions in Medicine (DICOM) for image transfer.
Selection of the most appropriate standard is important to enable simplified, quick, and reliable connections to a wide range of heterogeneous instruments, e.g.
laboratory, radiology, echocardiography, and electrocardiogram. The choice of Figure 4. Data exchange using HL7-IE gateway for laboratory tests between the HIS and LIS
Message Quantity (No. & Bytes) .
3 43 14 21 92 0 22321 20581 6 50 78 20744 21226 25 06 5 7 89 17 35653 14721 22 91 6 23020
Laboratory Data Exchange Using HL7-Interface Engine
The Journal on Information Technology in Healthcare 2003; 1(5): 331–342 337 standard should ideally be well-established, widely used, durable and allow for future expansion.
For the reasons given above we chose to use HL7 standards for data exchange, including replacing our own NTUH/LICI standard with the HL7-IE to manage and route messages. Using these standards we successfully managed to connect three automatic biochemistry machines to the LIS and HIS systems in our hospi-tal. Currently, the system is running smoothly and we now plan to extend this approach to other heterogeneous medical instruments at NTUH. Based on our Figure 5. Decision tree to help decide if the HL7 standard should be used to connect medical instruments with laboratory and hospital information systems
The decision tree is based on our planning, decision and implementation experience with HL7 at NTUH.
Is current environment heterogeneous?
Are systems modulated? Is future environment heterogeneous?
Cheng, Lai, Shyu et al
338 The Journal on Information Technology in Healthcare 2003; 1(5): 331–342
experience we have produced a flow chart (Figure 5) to help chief information officers (CIOs) decide if they should implement HL7 to connect MIs to the LIS and HIS in their organisation.
The approach we have taken has worked well in our hospital with the three connected Toshiba instruments. However, before we can recommend other medi-cal institutions adopt it, further work is required to establish its applicability and reliability with other heterogeneous MIs. Planned work includes:
1. Connection of other laboratory medical instruments.
2. Connection of other medical instruments, e.g. electrocardiogram, ultra-sound, and x-ray machines.
3. Identification of the best HL7-IE gateway backup solution.
4. Improvement of the speed-up issue.
5. Implementation of a security module for Internet data exchange.
6. Introduction of a variable-length message passing method to replace the fixed-length one.
Finally we would like to quote Stanley Huff, a senior medical informaticist at International Health Care, Salt Lake City, Utah and the previous chairman of the board of directors of HL7: “If CIO’s know how to implement standards, then they could reduce the costs and improve the speed of installation of electronic medical records systems.”17 We believe the work in this paper confirms the value of this statement.
ACKNOWLEDGEMENT
This research result is part of the “Research for Deriving an Object-Oriented Electronic Health Record Data Processing Platform” project which is supported by the PhD Student Research Grant (NTUH.92-M014), National Taiwan University Hospital. The NTUH/LICI standard has been supported by XEVOX (Tai-Eo Lai and Ben-Kun Shu). The NTUH HL7-IE solution has been supported by the Information Systems Department (Jung-Lin Wu, Yu-Fang Wu, Chi-Hwa Chen, Mam-Chan Chang, Chun-Li Niu, Yu-Shiang Weng, Ming-Tsung Lai and Chia-Ping Tsai), Laboratory Department (Dr. Wen-Chen Cheng and Dr. K.S. Tsai), Toshiba Taiwan Co.
(Joe Hong), TECO Inc. (Shih-Chan Fan), and Inqgen Co. (Bai-Hsiang Cheng, Shih-Hung Huang, Louisa Ho, Yu-Na Chiu, and Wei-Da Huang). The authors are grateful to all the above persons for their collaboration and contribution to this article.
REFERENCES
1 Li YC, Kuo SS, Jian WS, Tang DD, Liu CT. Building a generic architecture for medical information exchange among healthcare providers. International Journal of Medical In-formatics, 2001; 61: 241–46.
2 Cheng PH, Shyu FM, Lai JS, Chen SJ, Fan SC. e-Hospital on demand: medical application services architecture (MASA), Proc. of Medical Informatics Symposium in Taiwan, Taipei, Taiwan, 2002; 1: 45–50.
3 Arden Syntax. http://www.hl7.org/ and http://www.cpmc.columbia.edu/arden/.
Laboratory Data Exchange Using HL7-Interface Engine
The Journal on Information Technology in Healthcare 2003; 1(5): 331–342 339
4 Clinical Context Object Working Group (CCOW). http://www.hl7.org/.
5 Health Level Seven (HL7). http://www.hl7.org/.
6 Health Level Seven Taiwan (HL7-Taiwan). http://www.hl7.org.tw/.
7 American Society for Testing and Materials (ASTM). http://www.astm.org/.
8 Digital Imaging and Communications in Medicine (DICOM). http://medical.nema.org/
dicom.html.
9 Systematized Nomenclature of Medicine (SNOMED). http://www.snomed.org/.
10 Logical Observation Identifiers, Names and Codes (LOINC). http://www.regenstrief.org/
loinc/loinc.htm.
11 Common Object Request Broker Architecture for Medicine (CORBAmed). http://
www.omg.org/homepages/corbamed/.
12 ActiveX for Healthcare. http://www.activex.org/.
13 The Accredited Standards Committee (ASC) X12. http://www.x12.org/.
14 Linthicum DS. B2B Application Integration: e-Business-Enable Your Enterprise, Boston, MA:
Addison-Wesley, 2001.
15 Dick RS, Steen EB, Detmer DE. The Computer-Based Patient Record: An Essential Technol-ogy for Health Care, revised edn., Washington, DC: National Academy Press, 1997.
16 Chang KH, Lai JS, Shyu FM. Establish the EMRs using HL7 – experience on LAB system and ECG system of NTU. Proc. of Asia-Pacific HL7 Conference on Healthcare Information Standards 2002; 1: 75–89.
17. McCormack J. Standard: standard bearers cite progress, Health Data Management 2000; 8:
98–102.
APPENDIX A
Technical details of the data exchange between MIs, LIS and HIS at NTUH.
1. A clinical doctor keys in laboratory orders from HIS at an IBM mainframe 3270 terminal. A laboratory technician confirms the specimen is received and logs in the laboratory order entry at HIS. HIS then calls the IBM-HP3 program to pre-process the order entries into an HL7 (ORU^R01) message and passes this message to LIS at HP-370D. LIS uses LIS-HP3, a PowerBuilder program, to receive this HL7 message automatically. After that, LIS-HP3 is called to retrieve the HL7 message, to parse it and to call another database stored procedure (LIS-SP3) to store the related informa-tion in our LIS database.
2. LIS transmits the information for a specific order item to a specific MI. For example, LIS can call a PowerBuilder program (LIS-HP6) to group the HL7 (OML^O21) message and send it to a Toshiba biochemistry instrument.
This Toshiba biochemistry instrument will use a Visual Basic program (TSB-HP6) to receive and parse the HL7 message and pass it to corre-sponding medical instruments.
3. The specific MI processes the specimen, generates some reports, and then sends these reports to LIS. For example, the Toshiba biochemistry instru-ment will call a Visual Basic program (TSB-HP7) to create an HL7
Cheng, Lai, Shyu et al
340 The Journal on Information Technology in Healthcare 2003; 1(5): 331–342
(ORL^O22) message which contains the laboratory report and to send this message to LIS. Then LIS will use LIS-HP7, a PowerBuilder program, to receive and parse the HL7 message. Finally, LIS-HP7 will help to save related laboratory reports into the LIS database.
4. When a laboratory technician or a doctor wants to release the laboratory reports to HIS, LIS can call a PowerBuilder program (LIS-HP4) to collect the report, to group it into an HL7 (ORL^O22) message, and to send this message to HIS. HIS uses a corresponding IBM COBOL program, IBM-HP4, to receive and parse this HL7 message, then to save this report to the IBM IMS hierarchical database.
APPENDIX B
Comparison between the new HL7-IE gateway and the original Sybase gateway solution from the viewpoint of laboratory medical instrument connections
We used several different programming languages to achieve our HL7-IE gate-way solution. These were IBM COBOL, PowerBuilder, and Visual Basic. We also tried other programming languages to achieve system integration, including Java, and JavaServer Pages (JSP). We would recommend that newly created systems run in a Java platform.
HIS side (IBM mainframe)
LIS side
MI side
Flow 1: Sybase gateway is used to support NTUH’s own-defined data exchange mes-sages and standard, LICI.
1. The COBOL programs for the Sybase OpenClient and OpenServer programs are hard to write.
2. The codes are hard to debug.
1. The PowerBuilder programs and the other open system programming lan-guages are almost the same at both so-lutions.
2. Have to write codes to group and parse the NTUH’s own-defined data exchange messages.
1. Use NTUH’s own-defined LICI data ex-change standards.
2. Limited speed. One message per sec-ond at most (60 per minute).
Flow 2: HL7-IE gateway is used to support
Flow 2: HL7-IE gateway is used to support