Informatics Methods to Enable Patient-centered Radiology
Patient-centered radiology is a topic of increasing importance in radiology, motivated by both physicians and patients. From the physician perspective, patient-centered radiology is being driven by the rapidly evolving medical and molecular knowledge that promises personalized medicine. Patients are also driving patient-centered radiology, with increasing expectations of accessing and reviewing their medical information and participating more actively in their care. Patient-centered radiology practice is crucial, affecting the visibility of radiologists in health care.
The patient-centered radiology paradigm brings important challenges to both physicians and patients. From the physician perspective, two important challenges are selecting the appropriate imaging procedure for specific patients and rendering the best imaging interpretation that is personalized to the particular patient (eg, by incorporating historical and clinical data). In selecting an imaging procedure for specific patients, radiologists have an increasing number of choices. There are many new diagnostic agents and imaging techniques, increasingly tailored on the basis of patient-specific information. Radiology is evolving from generic imaging protocols for broad categories of indications to specific, often complex, image acquisition protocols designed to answer specific clinical questions. Radiology procedures and the selection of imaging agents will be increasingly customized according to each patient's underlying or suspected disease process, particularly in the era of molecular imaging.
Rendering the best patient-specific imaging interpretation is another requirement for radiologists in the patient-centered radiology paradigm. The pace of discovery in radiology is rapid, and radiologists are challenged by the information explosion and their ability to keep pace with the latest knowledge that could affect their imaging interpretations. Tools are needed to help radiologists access and use current knowledge, to guide them in customizing imaging to each patient, and to help them render the most accurate diagnoses.
From the patient perspective in patient-centered radiology, the expectation is that they will assume active roles in their care, particularly being informed about the results of imaging procedures and participating in medical decision making after receiving those results. Patients, like physicians, are overwhelmed by the amount of information related to imaging available online from diverse sources of questionable validity, and they are looking for help. Informing patients about the diagnostic imaging options available as well as the results of their studies is a crucial way radiologists can help patients to be engaged in their care as well as to ensure that critical results are communicated to them. This is a particularly important opportunity because the volume of information in physician practices and the lack of systems to manage the flow of information sometimes result in delayed communication of results from referring physician to patients. Radiologists can adopt informatics methods to manage the information glut and can help communicate results to patients. At the same time, radiologists will become more visible in the care process and make patients aware that radiologists serve an integral role in patient care.
A second aspect of the patient perspective in patient-centered radiology is enabling patients to participate in the medical decision-making process. Just as personalized medicine is changing how radiologists will approach interpreting images, shared decision making personalizes the patient management process. In many cases, the results of imaging are suggestive, but not conclusive, for disease. The choice of next steps (additional imaging, biopsy, or watchful waiting) often is affected by patient preferences. Shared medical decision making permits patients to weigh the trade-offs in their utility of the different outcomes, the likelihood of each outcome, and their risk tolerance.
At the heart of the challenges for patient-centered radiology from both the physician and patient perspectives is accessing and exploiting large amounts of information. Physicians need to access current radiologic knowledge to customize imaging and accurately interpret results. Patients need access to radiologic results in a timely fashion and in assessing the likelihood of different outcomes given these results. Informatics provides methods that can enable patient-centered radiology by helping physicians and patients manage and use knowledge: how it is acquired, used, and moved around. From the physician perspective, informatics can enable personalized medicine by delivering guidance for examination appropriateness specific to patients and by providing diagnostic decision support in image interpretation. From the patient perspective, it can engage them more closely in their care by communicating the results of imaging procedures and by enabling shared decision making.
Diagnostic decision support for interpretation
After selecting the best radiologic procedures to be performed, patient-centered radiology requires that the most accurate interpretation be rendered given the clinical context and observed radiologic findings. Radiology interpretation includes three key steps: (1) perception of image findings, (2) interpretation of those findings to render a diagnosis, and (3) decisions and recommendations about patient management (next tests or treatments). Perception of image findings is clearly important in radiology, and many informatics methods collectively called computer-aided detection have been developed in several domains of radiology. Detecting image features is approached in a similar manner in all patients. On the other hand, interpretation and decision making depend heavily on individual patient characteristics, and these tasks should be supported in patient-centered radiology.
The need for assisted interpretation and decision making is underscored by the fact that there is a great deal of variation among practitioners in these skills. Studies have documented interobserver disagreement in the interpretation of imaging studies. For example, in mammography interpretation, there is substantial variation in sensitivity, specificity, and area under the receiver-operating characteristic curve among radiologists. Much of the variation in practice likely results from the complexity of processing the vast amounts of knowledge needed to interpret the myriad findings observed in imaging. Most of medical practice is not quantitative but based on "heuristics" to guide physicians on the basis of rules of thumb. Such heuristics can fail in a variety of circumstances when combinations of features related to diagnosis do not fit the expected patterns and practitioners do not recognize the impact of such circumstances.
Diagnostic decision support is an informatics method to help radiologists with image interpretation and decision making. Unlike computer-aided detection, these systems do not focus on detecting a finding of interest but rather focus on the decision process, specifically on the diagnostic reasoning process in medical applications. Decision support systems begin with the observations (the findings detected on images) and integrate them with a formal model of the knowledge used to make decisions, outputting the most likely diagnoses. Decision support systems incorporate patient-specific knowledge, such as clinical histories or results of other tests. This patient-specific information, in combination with the imaging findings, can enable these systems to help radiologists render better care that is personalized to patients and reduce variation in practice.
In most decision support applications, the computer uses a model of disease (such as a Bayesian network) that incorporates image features and clinical data as variables in the model. The model relates image features and other observed clinical parameters to the likelihood of each type of disease. By entering observed clinical and imaging features into the model, a post-test probability of disease can be calculated and used to guide decision making and patient management. Such decision support applications thus inform radiologists about the most likely diseases for particular patients.
The input to decision support applications is usually a controlled terminology-encoded structured radiology report, that conveys the imaging observations and the key clinical information to the decision support application. The decision support application uses its model of disease to deduce the best course of action and reports that information to the practitioner. For example, a decision support system for mammography was recently created to inform radiologists as to the likely diagnosis and to provide guidance regarding biopsy recommendations of suspicious lesions. A radiologist enters the findings using a structured reporting form, and information in that form is processed by the system to produce its output (in this case, a differential diagnosis, ranked by probability of disease;).
Discussion
The core aspects of patient-centered radiology that informatics can support include physician-specific activities (selecting the appropriate imaging procedures for particular patients and rendering the best imaging interpretation that is personalized to each patient) and patient-specific interests (communicating the results of imaging procedures and shared decision making). These share in common the need to manage large amounts of complex information, a task that is difficult for people without computer assistance. In fact, in current unaided practice, communication sometimes breaks down, decisions can be made without access to all the necessary information, and there is variation in physician practice.
The goal of informatics is to bring the relevant medical knowledge to physicians and to patients to inform and to help radiologists provide the best care on the basis of all patient-specific information and to engage patients in their health care delivery. These goals are well aligned with the objectives of patient-centered radiology. A few core informatics methodologies are common to several patient-centered radiology applications. The first methodology is controlled terminology. Creating and adopting a constrained list of terms in radiology is crucial for clear communication of imaging results to referring physicians and patients and to permit computer applications to process radiology information unambiguously.
The second informatics methodology is decision support. A vast amount of radiology knowledge currently disseminated in published articles can be codified into computer models that can provide direct diagnostic guidance to radiologist for each patient on the basis of patient-specific characteristics. Such applications are the essence of translational medicine, by applying the knowledge in the literature to transform practice.
A third key informatics methodology is the electronic representation and transmission of information. Radiology is now an all-digital discipline, yet much of the work flow continues to adopt paper-based and film-based practice. Such work flow is inefficient, particularly with respect to communicating the results of imaging studies to referring physicians and patients. Patients are increasingly accessing their medical records online, though images are generally not part of these systems. In the future, imaging will become part of online health information systems, streamlining the dissemination of imaging results to patients and referring physicians, and more closely engaging patients in their care.
Although the informatics methods discussed in this work are promising in enabling patient-centered radiology, there could be challenges in adopting these methods. In fact, few of the informatics methods described here are widely adopted among health care institutions. The first challenge is that commercial products vary in their features and ability to deliver the functionality required. Most existing hospital systems focus on storing and retrieving patient-specific information, but few incorporate additional knowledge such as appropriateness criteria or decision support models. It is often possible to add this functionality by working with the vendors, though this can be a long and costly process.
A final challenge to adopting informatics methods for patient-centered radiology is cost. Informatics systems, programming staff for in-house solutions, and consultation and support require investment, both financial and personnel. In the increasingly competitive radiology environment, such costs could clearly prove to be a valuable investment by enabling institutions to more readily provide patient-centered radiology services.
In the long-term, as standards and interoperability methods mature, health care information systems will likely evolve to provide the necessary informatics infrastructure for patient-centered radiology. Just as Internet technologies are now pervasive in most homes and integral to how people find and use information, in the future, informatics approaches and intelligent systems will also become pervasive in hospitals, being used routinely by physicians and patients as radiology becomes increasingly patient-centered.
Conclusion
Patient-centered radiology is an important emerging paradigm that brings challenges to radiologists and patients as well as opportunities for informatics. The core methods of informatics in patient-centered radiology help physicians and patients access the exploding amount of data, particularly imaging data, in a timely and efficient manner. Computer applications will soon be appearing to improve the clarity of radiology reports, to help deploy appropriateness criteria, to help radiologists make better decisions on the basis of imaging information, and to enable patients to participate in their care and in shared decision making. Informatics methods are not futuristic developments on the horizon; many applications are already in routine use in several institutions. Through the adoption of these methods for patient-centered radiology, the value radiologists bring to the care process will be heightened, and their role in patient care will become more visible to patients.