If indeed radiology is in the information business, then radiology departments nationwide have many challenges ahead, not only in rewiring and reconfiguring departments to capture and distribute radiological information, but in managing and analyzing the data? yielded. Decisions in Imaging Informatics gathered three physicians on the front lines of the offensive to shape radiological reports into usable data to participate in an electronic round table discussion. The participants were:
Jonathan Einbinder, MD, MPH, is assistant professor in the Department of Health Evaluation Sciences in the University of Virginia School of Medicine, Charlottesville. After internal medicine residency training at Brigham and Women’s Hospital, he completed a fellowship in
clinical computing at Beth Israel Deaconess Medical Center, receiving a master’s degree from the Harvard University School of Public Health.? Currently, he is Director of Data Administration for the University of Virginia Health System, Charlottesville,? and Project Director for the Clinical Data Repository, a relational data warehouse.
Curtis P. Langlotz, MD, PhD, is advisor for informatics and clinical trials, Biomedical Imaging Program, National Cancer Institute, Bethesda, Md; staff radiologist and adjunct assistant professor of epidemiology and computer information science, University of Pennsylvania Health System, Philadelphia; and president, eDict Systems Inc, Mount Laurel, NJ.
David S. Memel, MD, is the corporate vice-president of information management at PeaceHealth, a multistate nonprofit integrated delivery system in the Pacific Northwest. Memel is a board-certified, fellowship trained radiologist, with additional training in clinical effectiveness research from the Harvard School of Public Health. He is currently completing a master’s degree in medical informatics from the University of Utah School of Medicine. Prior to his position at PeaceHealth, Memel worked at Intermountain Health Care in Salt Lake City, Utah, and practiced as an academic radiologist.
IMAGING ECONOMICS: How will radiology reporting be different 10 years from now?
LANGLOTZ: In just a few years, reporting methods will change dramatically from what they are today. Almost all of us now use conventional dictation and transcription to create text reports. Another small group of radiologists uses speech-recognition systems. In a few years, almost all reports will be structured reports containing images, terms, and speech, created by special-purpose software and delivered over the World Wide Web. The text report will be a mere by-product.
One early indicator of this trend is the recent approval of the Digital Imaging and Communications in Medicine structured reporting (DICOM-SR) standard, which will lead to widespread changes in radiology reporting methods. Among other things, DICOM-SR outlines standard methods for linking portions of a report (including terms and codes from standard lexicons) to specific locations on images. It will not be long before we see structured reporting systems that enhance radiologist work flow, provide decision support for radiologists, create multimedia reports, and transmit coded report information to an electronic medical record. These features will provide strong incentives to adopt structured reporting (by making the radiologist’s job easier while saving transcription costs).
MEMEL: Within 5 to 10 years, most radiology will be practiced in a real-time mode. Reporting will be done almost entirely, if not completely, using speech-recognition systems, with or without structured reporting. I also believe that natural language understanding systems will be integrated within the speech-recognition systems, thus facilitating automated, point-of-care medical-record encoding. This process will facilitate the successful use of real-time clinical decision support. In addition to speech recognition and natural language understanding, it is likely that, within 10 years, computer-aided diagnosis systems will be sophisticated enough that they, too, will be integrated into the reporting process.
EINBINDER: Actual images, rather than text reports, will be increasingly available in a distributed fashion. This will lead to changes in work flow, consultation, and communication with ordering clinicians (and patients). Along with distributed availability of images, I expect an increase in structured reporting, speech recognition, and natural language processing. The goal of these changes will be to make images (and data) widely available and usable while maintaining radiologist efficiency. Networking and image transmission will make it possible for radiologists to interpret studies off-site. Institutions and providers may choose to contract with geographically distant providers or consultants. Those who provide rapid, high-quality data and consultations in a standard, usable form will have an advantage.
IMAGING ECONOMICS: How will the electronic medical record affect the acquisition and storage of imaging data?
EINBINDER: The next generation of electronic medical record systems will begin to support high-fidelity image display on conventional hardware. In addition, the types of images contained in the record will expand to include static radiology studies (radiography, CT, and MRI); video (ultrasound and angiography); and other images (pathology, cytology, and photography). Already, electronic medical record and radiology vendors are introducing support for multimedia use. Radiology departments will need to capture, store, and transmit images in ways that can be used by these electronic medical records. A necessary prerequisite for the multimedia electronic medical record will be wide acceptance and adoption of standards for image storage and transmission, as well as development of the ability to transmit and display diagnostic-quality images on conventional workstations and displays rapidly.
LANGLOTZ: The move toward adoption of electronic medical records will increase the push for structured reporting. Right now, other hospital departments are content to receive radiology reports in text format. As electronic medical records become more pervasive, however, there will be increasing pressure for radiologists to provide coded and structured reports that are compatible with the sophisticated coding, searching, and matching algorithms of electronic medical records.
MEMEL: At a higher level, electronic medical records could affect the practice of radiology in general. The impact on the practice of radiology of electronic medical record systems implemented in areas outside radiology departments will depend on the functionality of the electronic medical record, as well as on the implementation and integration of radiology information systems (RIS) and electronic image management and communication systems with the electronic medical record.
One example of the way in which functionality can affect the impact of the electronic medical record on radiology practice would be physician order entry. If physician order entry is implemented in the electronic medical record, it could provide for the use of concurrent clinical decision support in the ordering of imaging studies. Through the use of physician order entry, clinical practice guidelines can be made available to the ordering clinician, potentially improving appropriateness in the selection of imaging studies.
The impact of an electronic medical record on the way in which medical imaging data are acquired and stored will depend on the synergy among the electronic medical record, the RIS, image-acquisition equipment, and the electronic image management and communication system. If well integrated, information from the electronic medical record could be used to support patient registration, appointment scheduling, work-list management, automated retrieval of prior imaging studies for comparison with new studies, and routing of images and reports to the appropriate radiologists and referring clinicians.
Most images will be stored in a digital format in the next 5 to 10 years. In order for the digital storage of images and the integration of RIS, electronic image management and communication systems, and electronic medical records to add value to the clinical process, it will be important to have integrated workstations for viewing patients’ clinical information, images, and imaging reports.
IMAGING ECONOMICS: To what extent will radiologists be asked to change their current work patterns to improve data capture?
EINBINDER: Throughout health care, there is increasing recognition of the need to capture data in a usable form for decision support, guidelines, retrospective analysis, and reimbursement. Other clinicians are already changing their work patterns, most notably for compliance with coding regulations and payor requirements. Actual patient-care interactions and documentation practices have been the most resistant to change, but change is occurring.
Compared with other specialties, radiology has a potential advantage: in many instances, content and work flow are already structured, repetitive, and heavily formatted. For example, at the University of Virginia and other institutions, pathologists and radiologists may generate reports (such as those for mammograms and Papanicolaou tests) in coded form. I expect the use of coded data capture and reporting to increase dramatically, driven by the need for usable data and by the potential increases in efficiency and report completeness made possible through the use of macros and templates.
An alternative approach to increasing coded data capture is parsing free-text reports to generate structured content. This approach would require fewer changes in work flow. I do not think that free-text parsing will be sufficient to meet the need for coded data, but it will be used in combination with the increased direct entry of coded information by radiologists.
LANGLOTZ: The need for cost savings and reduced report turnaround time has produced strong incentives to adopt speech-recognition systems. Consequently, more and more radiologists are asked to accept adverse changes in work flow caused by those systems. The accuracy of speech-recognition systems has approached the threshold of usability, but it takes significantly longer to generate reports using speech systems. An extra 20 to 30 seconds per examination may not sound like much, but multiplied by the 100 or more examinations that a radiologist might interpret in a day, it starts to become important.
These speech systems continue to be adopted despite work-flow problems, probably because the savings go directly to the bottom line while the costs (such as impeded work flow and decreased job satisfaction) are more difficult to measure. Structured reporting software that has the same features that Einbinder describes for mammography is now being developed for other radiology subdisciplines. Such software will tend to enhance radiology work flow while providing the same economic advantages as speech-recognition systems.
MEMEL: Many people perceive the product of radiologists who perform diagnostic studies to be the images produced. While the images are an intermediate product, the final product is the radiologist’s interpretation of those images, in the context of the clinical information obtained about the patient. In other words, the true product is information. In order for the information produced by radiologists to be of value in the care of patients, it must be received by the clinician requesting the information before decisions about the care of the patient have been made.
In general, radiology continues to be practiced in a batch mode. Multiple studies are performed, filmed, matched with prior studies, and then delivered to the radiologist for interpretation at a later time. The time has come for us to move into the practice of real-time radiology. Radiologists must be committed to changing the work flow in their practices so we can make the images, and our interpretations of them, available to the requesting clinicians immediately. This will require us to change the processes of our practices so that they support the review, interpretation, and dissemination of images and reports as soon as they have been completed. In some cases, we will need to make the information available to multiple clinicians simultaneously.
In order to practice real-time radiology, radiologists will need to embrace technologies such as electronic image management and communication systems, computer-based patient record systems, and speech-recognition systems. In addition, radiologists will need to acquire skill and comfort with interpreting images using soft copy, as well as with structured reporting.
IMAGING ECONOMICS: Hospitals will still have large repositories of text reports for some time. How can optimal use be made of them?
EINBINDER: Text reports, already stored in electronic format, represent a rich resource, if we can unlock the information contained within them. Options include making the text of the reports searchable by phrase or string. Note that query specificity will be limited, however. For example, a search for “infiltrate” would identify studies for which the report contains the phrase “no evidence of infiltrate.”
Text reports could be parsed to extract a set of coded findings. There are experimental projects under way to develop techniques for parsing free-text reports. This approach is not used widely (if at all) in commercial products. Text reports could also be parsed manually to extract coded findings. This is a labor-intensive and expensive approach, but it may make sense for specific situations or projects.
LANGLOTZ: Some of my past research involved the retrospective review of large numbers of radiology reports, and they were not pretty. There is often a lack of clarity (and even an intentional vagueness and ambiguity) in our reports. Even in retrospect, it is sometimes difficult to tell what meaning was intended. As a radiologist, I am as guilty of this as anyone. We need to express uncertainty, and we are working with a lexicon that is not often clearly defined. There is no question that these shortcomings of current reporting practices will limit our ability to make good retrospective use of text reports. In the near term, we will continue to be stuck with searching using combinations of words and phrases, and all of the problems that go with that; anyone who has ever done a search on the Internet will know what I mean.
IMAGING ECONOMICS: How can data warehouses be used to evaluate the effects of new technologies?
LANGLOTZ: Radiology databases certainly can help us measure the utilization of imaging technologies and, therefore, assess the economic effects of each technology on a radiology practice. True technology assessment, however, is much more difficult because it typically involves creating causal linkages between imaging results and patient outcomes. Without some clear structure in our reports, we have no automated way to determine the results of particular studies. As we produce more and more structured radiology reports, how-
ever, radiology data warehouses will become increasingly useful for technology assessment through the linking of a coded report with the electronic medical record.
MEMEL: The purpose of a data warehouse in health care includes the evaluation of the effectiveness of new technologies. Through comparative evaluations of the impact of proven and new technologies on the medical, cost, and satisfaction outcomes of various patient populations, the effectiveness of new technologies can be determined. Furthermore, the information in a data warehouse could be used to assess the impact of new technologies on radiology department resource requirements and productivity, both of which would be related to cost outcomes.
In the field of radiology, we are rapidly becoming more sophisticated in the performance of cost-effectiveness studies. Work on the impact of imaging studies on medical outcomes is still in its infancy, however, and little work has been done on the relationship between imaging services and patient satisfaction.
EINBINDER: Data warehouses can provide data for utilization tracking, outcomes analyses, before/after studies, and virtual case-control studies. It is unlikely that a rigorous, controlled study could be completely carried out solely from a data warehouse, but it is definitely possible to provide information for management decisions or for hypothesis generation. Data warehouses make it easy to identify members of a given population and characterize their care. By selectively supplementing warehouse data with manually collected information for a particular group of patients, more rigorous and valid studies can be performed.
IMAGING ECONOMICS: Can a data warehouse improve the quality of a radiology department’s products?
MEMEL: From my perspective, the purpose of a data warehouse in health care is to provide a source of integrated clinical, satisfaction, financial, and operational information that can be used to evaluate the services that we provide. The types of evaluation performed should include assessment of medical, functional, satisfaction, and cost outcomes; assessment of the effectiveness of clinical practice guidelines; and identification of the best clinical and operational practices. The results of these evaluations should then be used to determine where we should keep doing the same thing and where we should make quality improvement efforts.
In the practice of radiology, a data warehouse can be used to evaluate a number of clinical and operational processes and outcomes, with clinical evaluations including the accuracy of diagnoses, the frequency with which the radiologist’s report provides the clinical information being sought, the cost and clinical effectiveness of different imaging modalities in the context of different clinical conditions, and the satisfaction of patients. Operational evaluations could include the time elapsed between receiving a request for a study and performing it, the time elapsed between study performance and interpretation delivery, repeat-imaging rates, and lost-film rates.
EINBINDER: Currently, a RIS may be queried to determine the type and number of studies done, patient demographics, and (possibly) study results. For radiologists, a data warehouse can provide access to similar information about radiographic studies linked to other patient and financial information, such as laboratory results, diagnoses, procedures, outcomes, costs, and charges. Thus, a data warehouse can potentially help the radiology department understand how it fits into the overall patient-care process and can identify areas warranting more detailed analysis and study. n
Kris Kyes, technical editor of Decisions in Axis Imaging News, edited this round table discussion.