O o o

this problem in the 1980s. He promulgated the use of standards for the naming of laboratory tests and the reporting of their results (discussed in detail in Chapter 32). Until these standards are more universally used in health care, however, the construction of biosurveillance systems that collect laboratory data from hospitals will be time-consuming and expensive (the monitoring of laboratory test results and orders from national laboratory companies is far more feasible at present, as we discuss in Chapter 8).

When discussing data exchange with a hospital or other healthcare organization, the key questions to ask about the laboratory information system are as follows: Does your laboratory information system use SNOMED and LOINC? Do you send the results to an HL7-message router, a data warehouse, or a POC system? Is the microbiology outbound feed structured (i.e., not a free-text report intended for printing or display on a computer screen only)?

6.5. Dictation Systems

Dictation systems are a mainstay of clinical data recording in the hospital, emergency department, and outpatient settings. Clinicians often use dictation systems to record a patient's history, observations made during physical examination, progress notes, interpretations of radiological examinations, and results of postmortem examinations. Dictation systems can range in complexity from a single part-time transcriptionist working with a word processor to pools of transcriptionists using dedicated systems produced by companies such as Lanier. Although dictations are rich with clinical detail—including the patient's presenting complaint, the history of the illness, exposure information, vaccinations, vitals signs and physical findings, and diagnostic impressions—the data are recorded in English and are difficult for computers to understand. In addition, the time delay between dictation and transcription delays the availability of the data. Nevertheless, the value of the information is sufficiently high for both biosurveillance and medical applications that researchers in medical informatics have developed approaches to processing these data, which we discuss in detail in Chapter 17, "Natural Language Processing for Biosurveillance."

The availability of dictations for biosurveillance purposes is lower than that for laboratory data because not all transcribed dictations are stored electronically in databases and not all institutions route electronic versions of the transcriptions through an HL7-message router. Thus, even when dictations are available electronically, custom interfaces may need to be built to the database that stores the dictations.

When discussing data exchange with a hospital or other healthcare organization the key questions to ask about dictations are as follows: Do you have a dictation system that produces electronic copy? Does it provide an outbound interface (either HL7 or proprietary)? Do you send the dictations to the HL7-message router, the data warehouse, or a POC system?

Which of the many types of dictation are stored by the system, and with what time delay do they appear from the time of dictation?

6.6. Radiology Systems

Radiology departments were early adopters of IT, and today many practices manage the reports of examinations electronically. A radiology department performs radiographic, ultrasound, computed tomography (CT), magnetic resonance imaging (MRI), and other examinations of patients. The results of many of these examinations are highly relevant to biosurveillance. Clinicians use radiological examinations to diagnose infectious diseases (e.g., pneumonias). The reports include the diagnostic impression of the radiologist (e.g., "the combination of mediastinal widening and pneumonic infiltrate is consistent with pulmonary anthrax''). Unfortunately, in the majority of practices, the radiologist dictates his report and it is subsequently transcribed; thus, the reports can be delayed a day or more by the transcription process and are transcribed in English (or another language). Researchers in medical informatics have investigated methods for processing dictated radiology reports to extract information about patient characteristics, such as presence or absence of pneumonia (see Chapter 17) (Jain et al., 1996; Knirsch et al., 1998; Chapman and Haug, 1999; Hripcsak et al., 1999; Fiszman et al., 2000a).

We note that technical advances have made it possible for radiology information systems to store the images themselves in digital form, although the market penetration of this functionality is relatively low at present with the exception of imaging modalities such as CT and MRI, which are inherently digital. Rapid access to the images themselves, especially of chest radiographs, would be of value to both hospital infection control and governmental public health, especially during outbreak investigations. There is a trend to make such images directly available to physicians in hospitals through Web browser-based interfaces, and biosurveillance organizations could negotiate to obtain permission to access such systems or otherwise use this emerging capability.

When discussing data exchange with a hospital or other healthcare organization, the key questions to ask about radiology information systems are as follows: Do you have a radiology information system that stores reports? Does it provide an outbound interface? Do you send the reports to the HL7-message router, the data warehouse, or a POC system? Do you store images digitally (and which ones), and can your physicians access images by using the Web?

6.7. Pathology Information Systems

Pathologists examine bodily fluids, tissue specimens, and organs. Pathology information systems are more recent additions to healthcare computing because of the image-intensive nature of pathology practice (gross and microscopic examinations). Thus, market penetration is lower than that of laboratory or radiology systems. The data that may be useful for biosurveillance include both orders (for pathology tests) and their results. Access to the data in these systems is relatively difficult, as many systems function as free-standing applications used to generate printed reports. Integration with other hospital systems has not been as critical of a design factor as in the laboratory or radiology department, so biosurveillance organizations must either develop the integration functions or influence hospitals and system designers to "biosurveillance enable'' these systems.

6.8. Pharmacy Information Systems

Pharmacy information systems receive and process orders for medications such as antibiotics or antidotes for toxins, which are relevant to biosurveillance because they provide indirect evidence that the patient's illness may be caused by an infectious disease and even hint at the nature of that disease. In the vast majority of hospital pharmacies, physician orders are received on paper order forms, which pharmacists transcribe into the pharmacy information system; thus, there is a delay from the time that the physician expresses his or her understanding of the clinical problem in the form of an antibiotic order to the time that information is available electronically. The reliability of the information is extremely good, however, because pharmacists use expert knowledge and contextual information (available in the orders themselves, the pharmacy information system, and sometimes from review of the patient chart or contact with the physician) to validate the order before dispensing a medication.

6.9. Order-Entry Systems

Order-entry systems are computer systems that ward clerks or clinic staff use to communicate a physician's orders electronically to the laboratory and other departments in the hospital. If an order-entry system exists, it is a single point of access for information about orders for laboratory tests and medications, as well as orders that a patient be placed under respiratory isolation. Orders also may state a patient's diagnosis in an entry entitled "admission diagnoses.''

Ward clerks or clinic staffers enter orders promptly, so the information is available without time delay, except the time lag from when the clinician writes the order on a paper order form until the time when a ward clerk transcribes the order, a delay that is measured in hours at most. The reliability and accuracy of this information is high.

In less than 10% of hospitals, physicians enter orders directly into computers, eliminating the time delay and creating an opportunity for direct interaction and decision support of the clinician (Tierney et al., 1993; Bates et al., 1994; McDonald et al., 1994; Sittig and Stead, 1994; Frost and Sullivan, 2003). We discuss the potential of direct interaction with the clinician and decision support later in this chapter.

When discussing data exchange with a hospital or other healthcare organization, the key questions to ask about order-entry systems are as follows: Do you have an order entry system? What fraction of clinicians use it, and what fraction of orders does it capture? Does it provide an outbound interface? Do you send the orders to the HL7-message router or a data warehouse?

6.10. Point-of-Care Systems

A POC system is a hospital (or outpatient) information system that includes bedside terminals or other devices for capturing and entering data at the location where patients receive care (Shortliffe et al., 2001). Clinicians use POC systems to record directly details of patient encounters, to review information, and to order tests and other services. POC systems replace many functions of the paper chart and, in fact, are sometimes referred to as electronic medical records (although that term is used so loosely that we recommend that it is not used). Vendors sell POC systems specialized for diverse settings, including the emergency department, physician offices, hospitals, intensive care units, long-term care facilities, and home health care. POC systems even exist for prehospital care settings. Emergency medical units may use "ruggedized'' handheld computers in the field.

Depending on the POC system, a clinician may enter some subset of the data listed in Table 6.1. A clinical information system that has POC functionality has the potential to become paperless as each clinician, the laboratory, and the radiology department contribute to the collection of data about a patient. Advantages of POC systems to a hospital include quicker access to clinical information, the ability to communicate orders more quickly, elimination of the difficulties involved in reading the products of poor penmanship, and the ability to harness integrated decision-support tools such as electronic formularies, drug interaction warning databases, and electronic implementations of practice guidelines.

POC systems are the future of biosurveillance. A POC system facilitates the electronic capture of key diagnostic data (and usually in a computer-interpretable form rather than English). POC systems typically include decision-support capabilities (discussed in Chapter 13) that alert clinicians to potential drug-drug interactions and even suggest diagnoses. It is possible to program the underlying computer decision-support system to notice that a patient may have pneumonia and Gram-positive rods in a blood culture (an example of automatic case finding) and alert the clinician to consider a diagnosis of inhalational anthrax (and even report this suspicion automatically to a health department). High degrees of suspicion based on regional events can also be incorporated into the computer analysis. These capabilities are the reason that POC systems with decision support are the future of biosurveillance.

At present, most estimates of the market penetration of POC systems are in the single digits. The surgeon general offices of the nation's military services, when interviewed, were not aware of wide-spread use of POC systems in military facilities; if they are deployed, such deployments may be scattered in specific facilities. The VA, on the other hand, has high level of deployment of POC systems. Reasons for low market penetration include cost and reluctance by physicians and other providers to adopt these systems. A large multihospital organization may make a strategic decision to deploy POC system. Kaiser Permanente, in California, anticipates that its facilities will be operating an integrated POC system within five years.

In the United Kingdom, by contrast, POC systems are ubiquitous (Benson, 2002). The value of these systems for public health surveillance is illustrated by the rapidity with which the United Kingdom can potentially implement an anthrax surveillance strategy. By changing the decision support logic only, once, in a central location, the ability to detect postal workers presenting with influenza-like symptoms at the time of phone or physical presentation to any primary care physician in the country will exist.

Lazarus et al. (2002) has claimed that a POC system (a commercial product Epicare; Epic Systems Corporation, Madison, Wisconsin;http://www.epicsys.com) can be effective for purposes of public health reporting and bioterrorism early warning even if it serves only 5% to 10% of the population in a region being monitored. More research with POC-based surveillance is required to elucidate the relationship between the completeness of sampling of a population and the size of outbreaks of different diseases that can be detected.

6.11. Patient-Care Data Warehouses

Ralph Kimball (eminent data warehouse authority) defines a data warehouse as "a copy of transaction data specifically structured for query and analysis.'' Large hospital systems often build or purchase data warehouses specifically to integrate data from multiple information systems and multiple hospitals to provide clinicians with a consolidated view (often via a Web interface) of patient data (Figure 6.2). We refer to such data warehouses as patient-care data warehouses to distinguish them from data warehouses used for business or research purposes (Shortliffe et al., 2001).

When a patient-care data warehouse exists, it represents a point of integration of data that, similar to the HL7-message router, is a leverage point for biosurveillance. Data warehouses acquire data from other systems and transform data into a common format (e.g., data type, domain, unit of measure), and load the data into special data structures tailored for the intended use. In the case of patient-care data warehouses, the data are stored in structures that support rapid retrieval of the complete medical record of a single patient.

Transformation improves the accuracy of the data by removing duplicates from data sent to the data warehouse by laboratory or other systems. This process also may translate different hospital identification codes into a single canonical form, allowing data collected by different information systems to appear the same. Transformation is extra work that a biosurveillance organization would have to do if it were to access data directly from radiology and laboratory information systems.

If a healthcare organization has a Web-based interface to a patient-care data warehouse (as in Figure 6.2), it may be possible for a health department to negotiate with the hospital to provide its epidemiologists with access to the patient data on a need-to-know basis. Moreover, the access can be integrated within the health department's biosurveillance system. Figure 6.3 shows a sequence of screens from an early version of the RODS biosurveillance system (circa 2002) in which the user notices an increase in the number of patients with chief complaints of diarrhea, drills down to a line listing of cases, and then selects a case for which he wishes to see the patient record. After asking for authentication (a password and username issued by the healthcare system), the RODS system automatically takes him to the screen shown in Figure 6.2, which is the Web-based interface to the healthcare system's patient-care data warehouse. Note that Web browsers can remember login names and passwords, so the epidemiologist only must enter these credentials the first time he accesses the system, after which the transition to the patient-care data warehouse is seamless. During the months after the anthrax postal attacks, this function was used many times to do rapid investigations of spikes in syndrome data. A user could review approximately 40 patient charts per hour in this manner, which is several of orders of magnitude faster than conventional shoe-leather methods.

The availability of patient-care data warehouses in healthcare is low to moderate at present. Note that in many healthcare systems, a patient-care data warehouse will be a component of a more comprehensive "electronic medical record'' provided by a vendor. However, it likely will still have a Web-based interface that provides a consolidated view of a patient's "chart.''

In theory, it should be easy for a biosurveillance organization to interface with a data warehouse, which, if it exists, may represent a single point of integration that can provide data that have already been integrated and transformed. The key questions to ask when discussing data exchange with a hospital are as follows: Do you have a data warehouse? Is it for clinical care or archiving and business analysis? Is it part of a more comprehensive vendor system (and which one)? Many data warehouses now have Web-based interfaces. Although these interfaces are now being standardized, there are two competing standards. One is being promoted by Microsoft and the other by Oracle, with other vendors lining up on either side (or even both sides). So an additional question is Does your data warehouse support either XML/A or JOLAP?

6.12. Patient Web Portals and Call Centers

Two additional types of information systems are beginning to appear in health care. Call centers are facilities that receive

FIGURE 6.2 Web-based interface to a patient-care data warehouse. This data warehouse collects data from multiple information systems in multiple hospitals owned by the healthcare system to provide a consolidated view of a patient's medical history for a clinician. (Reproduced by permission from the University of Pittsburgh Medical Center [UPMC].)

FIGURE 6.2 Web-based interface to a patient-care data warehouse. This data warehouse collects data from multiple information systems in multiple hospitals owned by the healthcare system to provide a consolidated view of a patient's medical history for a clinician. (Reproduced by permission from the University of Pittsburgh Medical Center [UPMC].)

telephone calls from sick individuals who require information, triage, appointments, or immediate assistance. The staff fielding phone calls use information systems to document the calls, typically recording diagnostic information (e.g., reason for call, symptoms, and nurse assessment). In systems for which potential use for biosurveillance has been studied (see Chapter 27), the data included the practice guideline selected by the nurse to manage the call and the reason for call with timestamps, locations, and disposition.

The most extensive use of call centers in biosurveillance is the United Kingdom's National Health Service (NHS) Direct, which is a nurse-led telephone help-line that covers the whole of England and Wales. Data on the following 10 symptoms/ syndromes are received electronically from 22 call centers and are analyzed on a daily basis; cough, cold/flu, fever, diarrhea, vomiting, eye problems, double vision, difficulty breathing, rash, and lumps. Significant statistical excesses (exceedances) in calls for any of these symptoms are automatically highlighted

FIGURE 6.3 Sequence of screens from an early version of the RODS biosurveillance system. After noticing an increase in patients with chief complaints of diarrhea (top screen), the user drills down to a line listing of cases (bottom screen). The user selects a case to see the patient's medical record. After providing authentication (overlying dialog box), the RODS system automatically takes the user to the screen shown in Figure 6.2, which is the Web-based interface to the healthcare system's patient-care data warehouse.

FIGURE 6.3 Sequence of screens from an early version of the RODS biosurveillance system. After noticing an increase in patients with chief complaints of diarrhea (top screen), the user drills down to a line listing of cases (bottom screen). The user selects a case to see the patient's medical record. After providing authentication (overlying dialog box), the RODS system automatically takes the user to the screen shown in Figure 6.2, which is the Web-based interface to the healthcare system's patient-care data warehouse.

and assessed by a multidisciplinary team. The aim is to identify an increase in symptoms indicative of the early stages of illness caused by the deliberate release of a biological or chemical agent, or more common infections (Harcourt et al., 2001; Cooper and Chinemana, 2004; Cooper et al., 2004a,b; Doroshenko et al., 2004; Nicoll et al., 2004).

A second large project that involves call centers is the National Bioterrorism Syndromic Surveillance Demonstration Program, coordinated by Harvard Medical School and Harvard Pilgrim Heath Care (Platt et al., 2003;Yih et al., 2004). There is no single call center for the United Sates; therefore, this project seeks to recruit and integrate the call centers for cities, regions and ultimately the entire country.

Patient Web portals provide similar functionality but are basically self-service, much like Web-based airline bookings. The types of data collected by Web portals justify their inclusion in this discussion, despite their very low market presence. Patient Web portals have the potential to collect symptom level data as early as the day of onset of illness. Call centers, if patients are encouraged to use them early rather than waiting for illness to progress, have similar potential. The reliability and availability of such data have potential to be very high, especially if such services are designed from the ground up with the needs of regional integration of data for biosurveillance purposes in mind.

0 0

Post a comment