Data Sources
 
Implementation of Laboratory Order Data in BioSense Early Event Detection and Situation Awareness System

Haobo Ma,¹ H. Rolka,¹ K. Mandl,² D. Buckeridge,^3,4 A. Fleischauer,¹ J. Pavlin⁵

1CDC, Atlanta, Georgia; ²Children's Hospital Boston, Boston, Massachusetts and Harvard Medical School, Boston, Massachusetts; ³Palo Alto Veterans Health Care, Palo Alto, California; ⁴Stanford University, Stanford, California; ⁵Walter Reed Army Institute of Research, Silver Spring, Maryland

Corresponding author: Haobo Ma, CDC, 1600 Clifton Rd., NE, MS E-06, Atlanta, GA 30333, Telephone: 404-498-6283; Fax: 404-498-3225; Email: HMa@cdc.gov.

Disclosure of relationship: The contributors of this report have disclosed that they have no financial interest, relationship, affiliation, or other association with any organization that might represent a conflict of interest. In addition, this report does not contain any discussion of unlabeled use of commercial products or products for investigational use.

Abstract

Introduction: Laboratory test orders constitute an early outbreak data source. CDC receives laboratory order data in HL7 format from the Laboratory Corporation of America (LabCorp) and plans to use the data in the BioSense Early Event Detection and Situation Awareness System.

Methods: These LabCorp data contain information on tests ordered and include the type of test ordered and the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM)-coded reasons for the order. A consensus panel was formed to group test orders on the basis of expert opinion into eight standard syndrome categories to provide an additional data source for early outbreak detection. A laboratory order taxonomy was developed and used in the mapping consolidation phase. The five main classes of this taxonomy are miscellaneous functional tests, fluid screening tests, system-specific tests, tests for specific infections (by primary manifestation), and tests for specific noninfectious diseases.

Results: Summary of numbers of laboratory order codes in each syndrome category are fever (53), respiratory (53), gastrointestinal (27), neurological (35), rash (37), lymphadenitis (20), localized cutaneous lesion (11), and specific infection (63).

Conclusion: With the daily use of laboratory order data in BioSense, the actual distribution of laboratory order codes in syndrome groups can be evaluated, allowing modification of the mapping.

Introduction

Laboratory test orders constitute an early outbreak data source because test ordering occurs earlier in patients' health service than laboratory test results or final diagnoses. Three basis exists for using laboratory order data for early outbreak detection. First, laboratory orders reflect the physician's assessment of the patient's condition and intent to confirm or differentiate diagnoses. Second, when laboratory order data feeds directly from national laboratories are used, it can provide good population coverage. Finally, laboratory orders in large hospitals or national laboratories are transmitted in the standard Health Level 7 (HL7) format electronically, which provides timely data collection and helps ensure data quality. Use of HL7 is increasingly important when constructing automated early outbreak detection systems to handle large data volume. In addition, standard laboratory terminology such as the Logical Observation Identifiers, Names, and Codes (LOINC) has been well adopted for laboratory orders in commercial labs (1). Using LOINC codes in laboratory order data improves interoperability and scalability for the detection system.

CDC's BioSense program receives laboratory order data from LabCorp, which operates a nationwide network of 31 primary testing locations, and approximately 1,100 patient service centers. It tests more than 340,000 specimens daily

Methods

Laboratory Order Data

CDC receives a daily transmission of HL7 laboratory orders from LabCorp. Data files are sent to CDC from LabCorp using the Public Health Information Network Messaging System (PHINMS) with digital certificates and encryption. Immediately upon receipt, HL7 messages are mapped to XML format and stored in a data warehouse.

The laboratory order data contain patient age, sex, ZIP code, timestamp information and LabCorp propriety codes and corresponding LOINC codes for each laboratory test ordered. Each laboratory message might have more than one laboratory order placed in one patient encounter. Each laboratory order has one or more subcomponents. For example, a hepatitis B laboratory order will include various antibodies and antigens as its subcomponents. In addition, International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) codes are used as reasons for laboratory orders. CDC BioSense does not receive laboratory order results associated with each order.

All data received are anonymous and cannot be traced back to the patient. Use of patient ZIP codes and year of birth are not sufficient for patient identification in the large population areas surveyed.

Consensus Panel

When the laboratory order data are identified, laboratory order codes are mapped into syndrome groups. The objective for the mapping was to provide useful information for early event detection.

A consensus panel of 19 persons was organized with expertise in surveillance, infectious diseases, and medical informatics to perform the mapping. Participants included representatives from the Council of State and Territorial Epidemiologists, the Department of Homeland Security, Harvard Medical School, the Johns Hopkins University Applied Physics Laboratory, Stanford Medical Informatics, the Walter Reed Army Institute of Research, the New York City Department of Hygiene and Mental Health, and CDC. The expertise of domain experts was used to ensure the validity and reliability of the mapping result.

Mapping Procedures

The 11 syndrome groups used in BioSense have been defined by a multiagency working group. The syndromes included fever, respiratory, gastrointestinal, lymphadenitis, specific infection, localized cutaneous lesion, rash, neurologic, botulism-like, hemorrhagic illness, and severe illness or death potentially caused by infectious disease. Detailed syndrome definitions are available on the CDC Emergency Preparedness and Response website (http://www.bt.cdc.gov/surveillance/syndromedef). For the laboratory test mapping, the working group determined that laboratory order codes for botulism-like illness, hemorrhagic illnesses, and severe illness or death do not exist; LabCorp data feeds and these syndromes were not included in the mapping.

Domain experts determined the mapping, which was performed in stages. At each stage, a subset of the panel worked on the mapping problem, and the rest of the panel reviewed the results and arrived at a consensus. Four group members volunteered for the first round of mapping. They mapped each laboratory order to one or more syndromes using a simple mapping form. For example, a laboratory order with the name of influenza A and B antibodies, quantitative can be mapped to respiratory syndrome group. After receiving the four sets of results, three panel members consolidated the results. During these processes, other panel members provided input and advice.

The syndrome mapping of each laboratory order code was combined and represented as numbers. The number in combined results reflected the number of experts' selections in that syndrome group for a specific laboratory order code.

To reduce variations in the mapping results, the following constraints were enforced on the mapping:

• Mapping results are consistent within the same LabCorp laboratory order panel or profile group.
• Mapping results are consistent with LabCorp's online documentation of its use and specimen.
• Laboratory order codes classified in the same disease or pathogen group in a laboratory order code taxonomy are mapped in the same way.

Laboratory Order Code Taxonomy

A taxonomy of laboratory order codes was created. The five top-level classes of this taxonomy are miscellaneous functional tests (e.g., Coombs test and HLA screening), fluid screening tests (e.g., antibody identification and cultures), system-specific tests (e.g., tests for respiratory or gastrointestinal systems), tests for specific infections by primary manifestation (e.g., Cytomegalovirus or Lyme disease), and tests for specific noninfectious diseases (e.g., Lupus test).

This taxonomy provides a mechanism for classifying laboratory order codes systematically and for identifying agreement in mappings. Grouping laboratory orders for the same disease or pathogen enabled easy determination of the convergence and variations in the mapping results.

Implementation

After grouping laboratory order codes and identifying variations, additional constraints were placed on the mapping to facilitate laboratory order surveillance in a production system.

• Two or more experts agreed on syndrome mapping. The consensus panel reviewed and approved the results.
• Mapping was consistent in three ways as specified previously.
• Specific considerations included that the fever syndrome category could be selected only if three experts agreed because fever syndrome is more general than the other syndrome groups; the specific infection syndrome was not selected if a laboratory order code could be mapped to other syndromes.

Data Management

CDC BioSense receives a large amount of laboratory order data from LabCorp facilities. Many HL7 messages are updates for the same laboratory order. Although all received messages are maintained in the data warehouse, the most recent message of the same laboratory order is flagged for use in the BioSense web interface and for data analysis.

Results

The consensus panel's first meeting was in May 2004. In June, the simple mapping table was created in ACCESS, and four experts volunteered to conduct the first round of mapping. Four mapping results were received and consolidated in July. The final result was reported to the working group in September 2004.

Of 309 laboratory order codes, 246 were mapped into eight syndrome categories (Table 1). Some laboratory order codes were mapped to more than one syndrome group. For example, laboratory orders for tuberculosis were mapped as both fever and respiratory syndrome groups. Gastrointestinal syndrome group contains 27 laboratory order codes (Table 2). Respiratory syndrome group contains 53 LabCorp laboratory order codes (Table 3).

In addition, of 309 laboratory order codes, 63 were regarded by the working group as nonapplicable in the syndrome grouping. For example, bleeding and coagulation laboratory orders, nonspecific orders (e.g., complete blood counts), and orders for specific purposes (HLA typing) were determined nonapplicable.

Laboratory order data have been used on the CDC BioSense surveillance production system since November 2004. Both internal CDC BioSense monitors and state and local BioSense monitors can review the data daily for abnormal patterns. The data management team in CDC BioSense has created deduplicated data files in SAS format to facilitate data monitors' queries of records.

Discussion

Laboratory order data might provide information that can facilitate early event detection. To reduce the number of categories under surveillance and enable the integration of laboratory orders with other surveillance data sources in BioSense, it is helpful to map laboratory order codes to syndromes before analysis.

To fully cover the received laboratory order codes, mapping was initiated with LabCorp's local codes. LabCorp has mapped most of its proprietary codes to LOINC codes. Results can be converted to the LOINC-based mapping in the future. In the mapping process, domain experts mapped LabCorp codes directly to 11 syndromes. A laboratory order taxonomy was applied in the consolidation phase, which helped reduce inconsistencies. Taxonomy also provides finer granularity as disease classifications for laboratory order codes. Disease-based classification will provide flexibility in surveillance.

This mapping process has several limitations. First, only four domain experts volunteered in the first round mapping. A larger group of experts could help the mapping results converge and reduce inconsistency. Second, some pathogens infect various organ systems. The precise mapping for their laboratory order codes could only be achieved when the specimen collected was also considered. For example, Haemophilus influenzae could be a cause of meningitis as well as pneumonia. If the specimen was collected from cerebrospinal fluid, this laboratory order could be mapped to the neurologic syndrome; for a sputum sample, it would be mapped to the respiratory syndrome.

LabCorp messages include both laboratory order codes and ordering reasons in ICD-9-CM codes. The correlation between laboratory orders and their order reasons should be explored. Because ICD9-CM code mapping has been defined, the correlation might help improve laboratory order mapping.

With daily monitoring of laboratory order data in BioSense, the actual use of laboratory order codes in syndrome groups should be evaluated, allowing modification of mapping. Certain issues in the use of laboratory order data require further study, such as avoiding redundancy when a series of laboratory orders, which can be grouped into the same syndrome group, was ordered on the same patient by more than one physician and defining an event when a patient received laboratory orders during a continuous period.

Conclusion

To implement the laboratory order data, laboratory order codes were mapped to syndrome groups by domain experts a priori. In addition, a laboratory order taxonomy was created to facilitate grouping laboratory order codes and identifying inconsistencies. Finally the data management team created deduplicated data for monitors' use.

Acknowledgments

Members of the working group included John Abellera, Council of State and Territorial Epidemiologists; Martha Davis, U.S. Department of Homeland Security; Kenneth Mandl, Harvard Medical School; Steven Babin, Joseph Lombardo, Stephen Jacob, Johns Hopkins University; David Buckeridge, Stanford Medical Informatics; Timothy Endy, Shilpa Hakre, Cara Olsen, Julie Pavlin, Walter Reed Army Institute of Research; Farzad Mostashari, New York City Department of Health and Mental Hygiene; Aaron Fleischauer, Samuel Groseclose, Mary Hamilton, Henry Rolka, Richard Hopkins, Daniel Jernigan, Haobo Ma, CDC. The authors also appreciate LabCorp for their data provision and their consultation availability. The authors also thank Roseanne English and Paul McMurray in the CDC's BioSense BioIntelligence Center for their data management support.

References

1. McDonald CJ, Huff SM, Suico JG, et al. LOINC: a universal standard for identifying laboratory observations: a 5-year update. Clin Chem 2003;49:624--33.

Use of trade names and commercial sources is for identification only and does not imply endorsement by the U.S. Department of Health and Human Services. References to non-CDC sites on the Internet are provided as a service to MMWR readers and do not constitute or imply endorsement of these organizations or their programs by CDC or the U.S. Department of Health and Human Services. CDC is not responsible for the content of pages found at these sites. URL addresses listed in MMWR were current as of the date of publication.

Disclaimer   All MMWR HTML versions of articles are electronic conversions from ASCII text into HTML. This conversion may have resulted in character translation or format errors in the HTML version. Users should not rely on this HTML document, but are referred to the electronic PDF version and/or the original MMWR paper copy for the official text, figures, and tables. An original paper copy of this issue can be obtained from the Superintendent of Documents, U.S. Government Printing Office (GPO), Washington, DC 20402-9371; telephone: (202) 512-1800. Contact GPO for current prices.

Date last reviewed: 8/5/2005