Status Report
Orius Software: Calculation of Rates and Epidemiologic Indicators, and Preparation of Graphical Output

Long On, Robert M Semenciw and Yang Mao

Abstract

Orius software produces disease surveillance statistics, including output for the Health Canada Web site "Cancer Statistics Online" and on-demand statistics at the Web site for the North American Association of Central Cancer Registries. It allows flexibility in accessing data; storing, modifying and customizing requests; and producing statistical results and graphical output. Results include age-standardized or age-sex-standardized rates, standardized incidence/mortality ratios (observed-to-expected ratios) in which one area is optionally selected as the reference, potential years of life lost, average annual percent change, Mantel-Haenszel and maximum likelihood rate ratios, and the comparative incidence figure.

Key words: age-standardized rate; cancer registry; software; vital statistics

Introduction

Orius software was developed in response to the need for descriptive epidemiologic statistics and graphical output for surveillance analysis and user requests. At present it is used internally by Health Canada staff for surveillance projects, including output for the Health Canada Web site "Cancer Statistics Online" <http://dsol-smed.phac-aspc.gc.ca/dsol-smed/cancer/index_e.html>, and to produce cancer incidence statistics on demand at the Web site for the North American Association of Central Cancer Registries (NAACCR), "CiNA+ Online" <http://www.naaccr.org/CINAPlus/index.html>. Common data sources include mortality, cancer incidence and hospital morbidity.

The flexibility in accessing data, storing, modifying and customizing requests, and producing statistical results and graphical output varies depending on the computing environment. Currently, the desktop version operates in a stand-alone mode on Windows 95/NT or in a client/server mode with the client component on Windows 95/NT and the server component on HP-UX (UNIX-based operating system). The Web-enabled version runs on the Windows NT platform and is compatible with any web server that supports the Common Gateway Interface, for example, Apache and Internet Information Server. This version can also be delivered in a cross-platform environment consisting of Windows NT and UNIX (HP-UX or Sun Solaris) servers.

Statistical Methods

The following statistics are currently produced with Orius software.

• Age- or age-sex-standardized rates and standardized incidence/mortality ratios (also called observed-to-expected ratios) in which one area is optionally selected as the reference^1,2
• Potential years of life lost³
• Average annual percent change (AAPC)
• Mantel-Haenszel and maximum likelihood rate ratios, and the comparative incidence figure⁴

The AAPC can be approximated by the slope of the log of the rates based on the following approximation, where y represents the annual rate and x the year.

Because the log rate is used, zero rates are excluded from the AAPC calculation, and hence a table is provided giving the number of available points. Although Poisson regression allows zero rates, it requires the creation and use of age-specific rates (or perhaps 10-year groups) when broad age ranges are involved, such as all ages. The extra effort required to implement this step is not part of the current menu system, but SAS code is available to produce the required file from standard output created by Orius.

Graphical Output

The Web-based version produces tabular statistics as well as bar charts, pie charts, line charts and chloropleth maps. ArcView GIS is used to generate these graphical outputs. The category colour for the maps is based on the rate quintiles. At present, graphical output is provided to the user in JPEG file format.

System Architecture

The software comprises three main components. The Manager Application creates the data definition that describes the available data sets. For each set of available data, a data definition package is created.

The Client Application (desktop version) accesses the choices available in the data definition package and presents the choices to the user through an index-tab-based menu. The user can create, modify, copy, delete and save the request selections in the local database stored on the user's machine, if desired. The Client Application also allows the user to submit the request to the third component SAS directly (in stand-alone mode) or through a TCP/IP socket connection (in client/server mode) to obtain the SAS statistical result file and ASCII labelled output print file. A middle-tier SAS server is required to manage multiple client connections and requests. It performs the actual SAS execution calls and returns output results to the client.

The Web version extends the existing architecture with the use of ArcView GIS to provide the graphical output as required. The desktop Client Application is replaced by a WebClient Application, which interacts with a user through a series of dynamically generated HTML pages to collect request information. Again, a middle-tier ArcView server is required to request graphical output services on behalf of the client. A SAS format library and/or user supplied formats provide information for the titles on output formatted for printing.

All components, with the exception of SAS and ArcView GIS, are written in VisualAge for Smalltalk. This product is an industrial strength, cross-platform, object-oriented software development environment. Within the environment is a comprehensive class hierarchy with many ready-to-use and reusable objects. It also contains an integrated development environment with features for browsing and writing user-defined classes and methods, incremental compilation, program execution and testing, and object inspection as well as a breakpoint debugger. Other integrated features include source code management, configuration management and support for rapid application development in a team setting. The statistical routines are written in the SAS programming language through the use of macro scripts. The graphical routines are written in Avenue, ArcView GIS's scripting language.

The following is a list of the menus.

1. Age ranges
2. Age midpoints
3. Standard populations
4. Geographic areas
5. Causes
6. Sex
7. Year range
8. Age standard population set
9. Data definition (using one or more lists in 1-8 above)
10. Data definition package (using one or more data definitions specified in 9 above)

On-line help files for the Manager Application and Client Application, accessed using standard browser programs, provide further information on the lists, file requirements, outputs and current program limitations.

Data required for the rate calculation are accessed through summarized tables created by a program or database views along with an optional index. These tables, described further in the on-line help menu, contain summarized outcome data by age, area, year, cause, sex and, optionally, race. For SAS tables, a direct point access option is used for retrieval efficiency.⁵ The required case and population data are selected separately and summarized according to information supplied for the request. The two resulting files are merged, rates are calculated, and finally the output listings and graphical output are created.

Discussion

Orius software provides output for typical requests within a few seconds (slightly longer for the Web-based version because of additional processing). In the client/server environment of the Public Health Agency of Canada at Health Canada there are normally only a few simultaneous users. Further experience with the NAACCR Web site will indicate how the program manages larger numbers of concurrent users in a Web environment. Using VisualAge for Smalltalk has resulted in a portable program with reliable and transparent communication for the user when it is used in a client/server mode. Analysts can access the SAS code when needed for additional requirements. It may be possible to add the graphics outputs available in the Web-based version to the current client/server version, but no development is planned. This addition would be particularly useful if results could also be supplied in the template of a common graphics program that would allow users to make further changes as required.

Acknowledgements

The following terms are registered trademarks of their respective companies.

References

1. Breslow N, Day N. Statistical methods in cancer research. Volume 2: The design and analysis of cohort studies. Lyon: International Agency for Research on Cancer;1987:48S81; IARC Sci Publ No 82.

2. Ury HK, Wiggins AD. Another shortcut method for calculating the confidence interval of a Poisson variable (or of a standardized mortality ratio). Am J Epidemiol 1985;122:197S8.

3. Romeder JM, McWhinnie JR. Potential years of life lost between ages 1 and 70: an indicator of premature mortality for health planning. Int J Epidemiol 1977;6:143S51.

4. Ferlay J, Black RJ, Whelan SL, Parkin DM. CI5VII electronic database of cancer incidence in five continents, volume 7. Lyon: International Agency for Research on Cancer; 1997; IARC CancerBase No 2.

5. Howard N, Pickle LW. Efficient data retrieval: direct access using the point option. In: Proceedings of the Ninth Annual SAS Users Group International Conference; 1984 Mar; Tallahassee, Florida. Cary (NC): SAS Institute, 1984:294S8.

Long On, Robert M Semenciw and Yang Mao, Cancer Bureau, Centre for Chronic Disease Prevention and Control, Public Health Agency of Canada, Health Canada

Correspondence: Robert Semenciw, Cancer Bureau, PHAC, Health Canada, Tunney's Pasture, Address Locator: 0601C1, Ottawa, Ontario   K1A 0L2; Fax: (613) 941-2057; E-mail: Robert_Semenciw@hc-sc.gc.ca

[Previous][Table of Contents [Next]