Berkowitz, Murray R. DO, MA
* The proposed medical informatics system must have the capability to perform data-mining tasks upon a geographic information system, demographic data, and a patient database built upon analyses of patient records of interest while ensuring patient confidentiality.
* Specification of evaluation data for documenting success of the proposed medical informatics system will require answers to questions about detecting system errors, determining consonance between symptoms and diagnoses, and detecting security breaches.
* Limitations on the medical informatics system are from a variety of sources, most notably the Health Insurance Portability and Accountability Act of 1996 and the federal Privacy Act of 1974, from “traditional” rivalries and noncooperation between various governmental agencies (eg, Department of Defense vs State Department) the various levels (eg, federal, state, local health departments), by such factors as resistance by physicians, nurses, and public health workers and by current reliance upon paper-based systems.
The terrorist attacks on the World Trade Center’s Twin Towers in New York City, the Pentagon near Washington, DC, and United Airlines Flight 93 on September 11, 2001 have brought the US population to full awareness that “it can happen here.” Since that time, the United States has been on heightened security alert and has attempted to become more capable of addressing potential future terrorist threats. Among these threats are the problems associated with biological weapons (bioterrorism). Given the many areas of this country that have high population densities, it is conceivable that the release of a biological weapon within one of these areas could have catastrophic effects. One of the key problems in combating bioterrorism is the identification of an event (ie, an attack) and then determining what biological agent or agents were released. Informatics has been proposed as a major component in the response planning, detection, and early-warning systems for bioterrorist attacks,1,2 and syndromic surveillance systems have been proposed as the main instrument in achieving these goals. Buehler concluded that syndromic surveillance has “advanced considerably,” but also recognized that its utility and recommendations for practice need to be developed.3 Mandl and colleagues analyzed syndromic surveillance systems and methodologies and identified areas for further research.4 Begier and colleagues showed that neurologic and sepsis syndromes had markedly lower agreement between the presenting chief complaint and discharge diagnosis than the other monitored syndromes.5 This is significant in that these syndromes will most likely be present in cases resulting from chemical or biological attacks. The current information systems for use in detecting bioterrorist attacks lack a consistent, overarching information architecture.6,7
Many pathogens have the potential to serve as biological weapons. The concerns revolve around the notions of releasing the disease agent into an unknowing and unsuspecting crowd. Each individual then becomes ill (a primary case). Worse still is that each infected primary case infects others in turn with whom he or she comes into contact (secondary cases), and the disease then continues to spread. For any agent to be appropriate for use as a biological weapon, it must be easy to carry and release, have a short incubation period, and be both virulent (have a high case-fatality rate) and highly contagious (ideally, able to infect others via human-to-human contact).
To combat potential bioterrorist attacks, primary prevention plays a major role. If the population is immunized against major or potential pathogens, then the effectiveness of any bioterrorist attack is thwarted or minimized. Because immunization against every possible pathogen is impossible, the next level of defense exists in the form of rapid identification of a bioterrorist attack, meaning that it must be recognized as having taken place, the specific pathogen(s) identified, and necessary treatment(s) administered.
An actual attack would most likely involve the surreptitious release of pathogens at a popular, widely attended event, such as a rock concert or sports event, where the agent is released quietly into an unknowing crowd. This would be most effective if the potential primary patients are densely packed and the agent can remain in a highly concentrated form. Clearly, an indoor event (eg, basketball game) would serve better than an outdoor event (eg, football game).
Each infected person (the primary cases) would return to his or her home, school, business, or similar location and infect others (secondary cases). This propagation would continue and the infection spread further. The primary cases would manifest as ill and seek treatment from their primary physicians, in urgent care facilities, or in hospital emergency departments. The number of patients examined would probably exceed some expected level of infection; that is, the treating healthcare providers would perhaps note an increase in the signs and symptoms usually associated with, for example, upper respiratory infections. The question then is whether this is a new strain of influenza, a new disease entity (eg, severe acute respiratory syndrome), or the result of some bioterrorist activity. Each of the several provider types will examine a varying number of such patients, and they may encounter a series of signs and symptoms that suggest that these patients have something more than a typical disease and notify the local, state, and/or federal health authorities; however, this likely will not occur. More likely, these patients will be (mis)diagnosed as having a “common” illness and they will be treated and released. With the use of an electronic medical record for each patient, data-mining algorithms could be searching for some predetermined level of illness or contagion (ie, “trip wire”) for each geographic region (eg, perhaps each postal zip code). These levels or trip wires would be set by government entities assigned the responsibility for maintaining the public health (eg, US Public Health Service, the various state departments of health and mental hygiene, local health departments). This is a political input threshold.
Once a bioterrorist attack has been confirmed, the pathogen must be identified so that appropriate and timely treatment can be initiated. Protocols for containing the infection and eliminating its spread must be determined and instituted.
Tang et al stated, “Effective public health practice and decision making depend on timely information, much of which is not available.”8 They advocated the National Health Information Infrastructure (NHII). William Yasnoff, the former senior advisor for the NHII to the Secretary of the US Department of Health and Human Services, has identified four domains of the NHII, one of which is public health informatics.9 Yasnoff and colleagues identified “bioterrorism, emerging infections and antibiotic-resistant organisms” as new challenges for the public health community. They reported the need for the development and implementation of a set of data standards at the heart of public health informatics10:
“Public health informatics, defined as the systematic application of information and computer science and technology to public health practice, research and learning, is the emerging discipline that integrates public health and information technology. The development of this field and the dissemination of informatics knowledge and expertise to public health professionals are critical to unlocking the potential of information systems to improve the health of the nation. Major challenges include developing coherent, integrated national public health information systems, increasing integration efforts between public health and clinical care system and addressing pervasive concerns about the effects of information technology on confidentiality and privacy.”
A national team of contributors has been developing and validating a model informatics system for its utility for public health.11 Yasnoff and colleagues stated, “A variety of educational and training programs to address public health informatics knowledge and skills are urgently needed by the public health workforce.”10 Koo and colleagues have assembled an overview of public health issues for clinical informaticists.12 They reiterate the goals and challenges of public health informatics espoused by Yasnoff and colleagues and add another set, as follows:
1. “Ensuring the capacity to assess community problems in a comprehensive manner through the development of integrated nationwide public health data systems
2. Facilitating the improved exchange of information between public health and clinical care
3. Privacy, confidentiality and security are pervasive and persistent challenges to progress in public health informatics
4. To apply information technology in unanticipated ways to reengineer public health and invent new ways to protect and promote community health”
Since the terrorist attacks on the United States in September 2001, followed rapidly by the anthrax (ie, bioterrorist) attacks through the mail in October 2001, the need for public health informatics systems has been demonstrated clearly. Syndromic surveillance systems have been developed and prototyped. The Real-Time Outbreak and Disease Surveillance system was prototyped by the Center for Biomedical Informatics at the University of Pittsburgh in 1999 and was more fully developed and implemented by a team from the Center for Biomedical Informatics at the University of Pittsburgh and the Department of Medical Informatics at the University of Utah and deployed at the 2002 Winter Olympics.13,14 The ESSENCE and ESSENCE II systems developed by the Johns Hopkins Applied Physics Laboratory serve as a disease surveillance test bed for the National Capital Region,15 and the BioWatch program serves as a more nationally focused syndromic surveillance system.16,17 Although the BioWatch program is reported to be the nation’s first early-warning system to detect biological attack,18 there have been many reported problems,19,20 such as false-positives.21,22 Notwithstanding the above, it has been reported that the underlying models used in the BioWatch program to predict the spread of infection may be inaccurate.23 The implementation of accurate stochastic models in predicting dispersal of infection is critical to the design and reliability of early-warning systems. Additional systems that can monitor the sales of over-the-counter products and detect increases in these sales can help to alert public health officials of possible outbreaks before those afflicted even seek professional care from their physicians, urgent care centers, or hospital emergency departments.24,25 These public health informatics systems demonstrate the need for professionals with competencies developed by the type of comprehensive public health informatics specialty training proposed by Yasnoff and colleagues.
Analysis, Specification, and Design of the Proposed System
System features include the development and implementation of a secure, inexpensive, universally available device for recording, transmitting, and receiving patient data and health treatment regimens and protocols. It must be easy to use (judged by 90% of users). The total development budget should be <$10 million and the operating budget should be <$4 million/year. The system should detect errors and correct them whenever possible and it must be able to detect security breaches. The system also must be able to determine the consonance between patient symptoms and appropriate diagnoses.
The sources of data for this system include clinical data from physician offices, urgent care centers, and emergency departments in electronic (eg, electronic medical record), voice (eg, telephone), and paper forms (eg, fax). The trip wire threshold is politically determined metadata. The sources of protocols for handling these cases are the appropriate health authorities at the local, state, and federal levels. Examples of these authorities include the county health office, state department of health and mental hygiene, Centers for Disease Control and Prevention, Food and Drug Administration, Environmental Protection Agency, Department of Health and Human Services, and Department of Homeland Security. Syndromic surveillance and environmental data provide other data sources and the use-case diagrams illustrate this (Fig. 1).
Given the complexity in analyzing the vast quantity of data and the large and varied number of parameters, a medical informatics system needs to be designed. The proposed system must have the capability to perform data-mining tasks upon a geographic information system, demographic data, and a patient database built upon analyses of patient records of interest.26 It is essential that patient confidentiality be ensured; consequently, it is imperative that no individual patient be identifiable via system query or by deduction from system analysis. These systems requirements are summarized as follows:
* Data mining
* Ensure total patient confidentiality
* Able to perform biostatistical analyses
* Able to perform epidemiological studies: measures of disease occurrence (risk, prevalence, incidence), medical surveillance of disease outbreaks (sensitivity, specificity, positive and negative predictive values), support clinical trials (including double-blind randomized trials), cohort studies, and case-control studies
The overall top-level information architecture for the proposed system is shown in Fig. 2. The data-mining module contains various algorithms to perform these tasks. Many advances in expert database systems, intelligent information retrieval, and processing algorithms have been made and the fruits of these advancements should result in a state-of-the-art subsystem. The biostatistical functions module contains the mathematical functions necessary to perform biostatistical analyses of the data-mining results.27 The collection of needed functions for this module is well established, and these data may be further analyzed by the application of selected epidemiological studies. The functions in this module must support measures of disease occurrence—risk, prevalence, and incidence—and must be able to support medical surveillance of disease outbreaks. Sensitivity, specificity, and positive and negative predictive values must be able to be determined. This module must support clinical trials, including double-blind randomization, and cohort and case-control studies also must be supported. Finally, a report generator (not depicted) aids in the presentation of the results, both oral and written. The functions inherent in this module are well known in the informatics community, and the specifications for these functions are elaborated in the accompanying functional specifications (Fig. 3).
Previous test beds for terrorism response and preparedness have shown the weaknesses of the current collections of “surveillance systems,” especially the lack of integration of the various system and communications modules.28 Specification of evaluation data for documenting success of the proposed system will require answers to a number of questions: How does the system detect errors? What is the error frequency? What is the degree of consonance between symptoms and diagnoses? How does the system detect security breaches? What is the frequency of attempted security breaches? What is the frequency of actual security breaches?
Limitations on the system emanate from a variety of sources. Some are the result of laws and regulations, most notably the Health Insurance Portability and Accountability Act of 1996 and the federal Privacy Act of 1974. There are limitations that result from “traditional” rivalries and noncooperation between various government agencies (eg, Department of Defense vs State Department) and the various levels (eg, federal, state, local health departments). Further limitations are imposed by factors such as resistance from physicians, nurses, and public health workers. Finally, limitations are imposed by the current reliance upon paper-based systems.
Future research and development directions will require the top-level information architecture, and functional specifications of the proposed medical informatics system undergo further decomposition into more detailed levels of systems analysis. Structure charts and data flow diagrams describing the more detailed levels must be determined. Entity-relationship diagrams will result from this further analysis and logical and physical design specifications may then ensue. The development and implementation of data-mining algorithms will be complex, and even implementation of well-established biostatistical and epidemiological functions will not be trivial.
1. Teich JM, Wagner MM, Mackensie CF, et al.. The informatics response in disaster, terrorism, and war. J Am Med Inform Assoc 2002; 9: 97–104.
2. Kohane IS. The contributions of biomedical informatics to the fight against bioterrorism. J Am Med Inform Assoc 2002; 9: 116–119.
3. Buehler JW. Review of the 2003 National Syndromic Surveillance Conference—lessons learned and questions to be answered. MMWR Morb Mortal Wkly Rep 2004; 53 (suppl): 18–22.
4. Mandl KD, Overhage JM, Wagner MM, et al.. Implementing syndromic surveillance: a practical guide informed by the early experience. J Am Inform Assoc 2004; 11: 141–150.
5. Begier EM, Sockwell D, Branch LM, et al.. The National Capital Region’s Emergency Department syndromic surveillance system: do chief complaint and discharge diagnosis yield different results? Emerg Infect Dis 2003; 9: 393–396.
6. Lober WB, Karras BT, Wagner MM, et al.. Roundtable on bioterrorism detection: information system-based surveillance. J Am Med Inform Assoc 2002; 9: 105–115.
7. Gilfillan L, Smith BT, Inglesby TV, et al.. Taking the measure of countermeasures: leaders’ views on the nation’s capacity to develop biodefense countermeasures. Biosecur Bioterror 2004; 2: 320–327.
8. Tang PC, American Medical Informatics Association Public Policy Committee. AMIA advocates national health information system in fight against national health threats. J Am Med Inform Assoc 2002; 9: 123–124.
9. Yasnoff WA. National Health Information Infrastructure: Key to the Future of Health Care. Rockville, MD, US Department of Health and Human Services, 2002.
10. Yasnoff WA, Overhage JM, Humphreys BL, et al.. A national agenda for public health informatics: summarized recommendations from the 2001 AMIA Spring Congress. J Am Med Inform Assoc 2001; 8: 535–545.
11. Public Health Data Standards Consortium. Electronic Health Record: Public Health Perspective. Public Health Data Standards Consortium Ad Hoc Task Force on Electronic Health Record–Public Health (EHR-PH); 2004.
12. Koo D, O’Carroll P, LaVenture M. Public health 101 for informaticians. J Am Med Inform Assoc 2001; 8: 585–597.
13. Tsui FC, Espino JU, Data VM, et al.. Technical description of RODS: a real-time public health surveillance system. J Am Med Inform Assoc 2003; 10: 399–408.
14. Gesteland PH, Gardner RM, Tsui FC, et al.. Automated syndromic surveillance for the 2002 Winter Olympics. J Am Med Inform Assoc 2003; 10: 547–554.
15. Lombardo JS. The ESSENCE II Disease Surveillance Test Bed for the National Capital Area. Johns Hopkins APL Tech Dig 2003: 24: 327–334.
17. Estacio PL. Bio-Watch overview. CBRN countermeasure technologies. Environmental bio-agent detection & biosurveillance initiatives. www.hsdl.org/?view&did=479074
. Published September 10, 2004. Accessed October 22, 2012.
19. Goldstein BD, DeSimone JM, Ascher MS, et al.. Effectiveness of National Biosurveillance Systems: BioWatch and the Public Health System. Washington, DC, Institute of Medicine/The National Academies Press, 2009.
20. Goldstein BD, DeSimone JM, Ascher MS, et al.. BioWatch and Public Health Surveillance: Evaluating Systems for the Early Detection of Biological Threats. Washington, DC, Institute of Medicine/The National Academies Press, 2010.
23. US House of Representatives, Committee on Government Reform, Subcommittee on National Security, Emerging Threats and International Relations. Following Toxic Clouds: Science and Assumptions in Plume Modeling. 108th Congress, June 2, 2003. http://www.hsdl.org/?view&did=695347
. Accessed November 28, 2012.
24. Wagner MM, Robinson JM, Tsui FC, et al.. Design of a national retail data monitor for public health surveillance. J Am Med Inform Assoc 2003; 10: 409–418.
25. Hogan WR, Tsui FC, Ivanov O, et al.. Detection of pediatric respiratory and diarrheal outbreaks from sales of over-the-counter electrolyte products. J Am Med Inform Assoc 2003; 10: 555–562.
26. Brossette SE, Sprague AP, Jones WT, et al.. A data mining system for infection control surveillance. Methods Inf Med 2000; 39: 303–310.
27. Burkom HS. Development, adaptation, and assessment of alerting algorithms for biosurveillance. Johns Hopkins APL Tech Dig 2003; 24: 335–342.
28. Simmons SC, Murphy TA, Blanarovich A, et al.. Telehealth technologies and applications for terrorism response: a report of the 2002 coastal North Carolina domestic preparedness training exercise. J Am Inform Assoc 2003; 10: 166–176.