Unlocking the Business-intelligence Vaults in Radiology

Twitter icon
Facebook icon
LinkedIn icon
e-mail icon
Google icon
In the current climate of accountability, regulation, and continual improvement, radiology managers and administrators are called upon to make timely and informed decisions that affect the quality of their departments’ output (the radiology report) and the financial viability of their practices. Successful managers need a system of measurements (metrics) to reflect the performance of their units in order to achieve strategic goals. In a radiology department, this translates to the manager’s need for access to live data demonstrating workflow and identifying bottlenecks, in addition to the need for on-demand fiscal reports. The time it takes to build a report using traditional methods reduces both the value of the information and the time that an organization then has available to take action to correct a deficiency. The traditional report can only answer questions that are identified prior to a periodic meeting, and there is no way to answer questions that might occur during the meeting. Without reports based on accurate, current, integrated data, decisions might not be based on an accurate picture of workflow and resource utilization. For example, the chief technologist might be requesting additional manpower, but without information on how the current workforce is being used, a decision might be based on anecdotal evidence. Perhaps the number of studies performed by the department might be increasing, but reimbursement is decreasing, and the manager needs to identify the reasons for the discrepancy. During a meeting, a question might come up regarding the utilization of a modality such as MRI, and if this question was not anticipated, a report might not be available. With on-demand reports generated by business-intelligence analytics, a report can be generated in real time. Business intelligence is loosely defined as using computer techniques and databases to identify and organize information about business processes and trends. Other terms are used in the radiology literature to define the process; these include data mining, data analytics, and using a digital dashboard. Typically, a digital dashboard refers to real-time measurements resembling the dashboard in a car that gives the driver indications of compliance (speedometer) and resource utilization (fuel, oil, and air in the tires), along with warnings of potential problems. Other business-intelligence analytics can provide static reports on the day-to-day operation of the department. The difference between these business-intelligence analytics and typical reports generated by various information systems is that business intelligence uses data from multiple sources and can generate these reports on demand. Business-intelligence tools and dashboards are described in detail in the literature or in presentations at meetings, but there is very little guidance for the manager who wants to use these tools. Identifying the Target Before setting out to find an individual solution for a business unit, the manager needs to define the problems that he or she wishes to investigate. A team approach will help identify specific targets for improvement. Digital dashboards or business-intelligence tools in the radiology literature usually focus on one or more of these general topics: resource utilization, workflow improvement, improvement in reporting, reimbursement, and quality assurance (to include safety). Resource-utilization analytics provide reports on the productivity of staff and the utilization of imaging modalities. Workflow dashboards can give the user a real-time display of the flow of patients and procedures through the department. Radiology-interpretation dashboards tend to focus on real-time views of report-turnaround times, and reimbursement models focus on the appropriateness of studies. Ensuring the appropriateness of performed procedures can lead to improvement in reimbursement. Quality-assurance analytics are used to evaluate errors and to monitor safety measures. For example, two of the Agency for Healthcare Research and Quality’s metrics deal with reducing radiation exposure, and the agency recommends recording CT dose and documenting fluoroscopic examinations’ duration.1,2 Although tools to measure and monitor radiation exposure are not yet widely available, these quality measures will become mandatory. Unfortunately, digital dashboards and other business-intelligence tools are not yet widely used, partly because radiology data are distributed across widely disparate system and partly because there is no plug-and-play commercial solution that will work with all systems. The institutions that report phenomenal successes with these tools either have a team of imaging-informatics professionals who have the time and resources to create a one-off solution3-13 or have a single-vendor solution in place that applies to all the medical information systems throughout the institution. The basis for a radiology business-intelligence system or solution is the set of data that probably already exists in an institution. This dataset includes patient information, radiology orders, radiology images, and reports. In addition, the institution will be most likely to have the required infrastructure, including the networks and servers that support the enterprise. Data acquisition is triggered by specific events, such as patient registration, order entry, performance of a study, or reporting of the study’s results. In addition to the basic data generated by the workflow (orders, images, and reports), each of these activities generates additional information on the timing, location, and personnel involved. If the organization is to build a business-intelligence solution, selected data must be integrated and stored in a repository for analysis. An analytics engine will mine the aggregate data in response to requests from users and will generate reports, which are then displayed through a user interface. The components of a business-intelligence system are shown in Figure 1 and include the usual sources of data for radiology applications: the hospital information system (HIS), the RIS, the PACS, and the dictation system.
image
Figure 1The components of a business-intelligence solution include sources of data for radiology applications, including the RIS, PACS, hospital information system, and dictation system.
The trick is to get all of the required data into a single consistent format in what is shown as the business-intelligence database, so the analytics engine can formulate real-time analysis, as well as historical data reports. This is not an easy task, and it will require work on the part of managers; they must identify their goals, gain cooperation among various IT sections, collaborate with vendors, and perhaps obtain outside help from consultants. A Recipe for Success The first thing that radiology managers and administrators will need to do is to articulate the goals that they wish to achieve using their business-intelligence solution, ranking them in order of importance. The table shows some of the more common goals and which systems might contain the data required to meet those goals. This is the time for managers to engage other stakeholders in the process. Stakeholders will include the radiologists, IT personnel, clerks, and technologists, each of whom will have their own individual concerns and priorities—and potential conflicts. For example, an IT manager might not be willing to assign resources to the work involved in identifying sources of data and might be reluctant to approve additional information systems to support the objective. Technologists might resist an effort to quantify their work product by the number of studies performed because it could affect their performance reviews. Managers will need to use tact and all of their leadership skills to show that these tools will benefit everyone by improving departmental performance and remuneration (and possibly patient satisfaction). Two business-intelligence solutions are widely described in the literature, probably because the data required for each are readily accessible, and the applications can help a department improve its total turnaround time (from the time that a patient enters the department until the report is communicated to the referring physician). Rapid turnaround appears to result in improved perception of value on the part of the referring physician, in addition to enhanced patient satisfaction.
image
Table.Business-intelligence Solutions and Data Sources
The illustrative scenarios that follow would be appropriate for a real-time, interactive dashboard application that could help managers identify bottlenecks as they occur, or they could be implemented as static, on-demand reports. The first application is a dashboard showing departmental workflow, with the ability to drill down to evaluate the suspected problem areas. The second application is a real-time view of dictation, signing, and communication of reports. Departmental Workflow A simple description of the process of performing a radiology study can be broken down into discrete steps defined as order entry, patient arrival, patient preparation (if necessary), imaging, verification that imaging has been completed and is correct, report generation, report signing, and report communication. As the table shows, order entry is performed using either the HIS or RIS, patient arrival is marked in the RIS, patient preparation might be entered in the RIS (especially if the preparation includes the administration of drugs), imaging and verification will be included in the PACS, and reporting will use the dictation system. It is probable that some imaging steps (such as study completion) will be reported to the RIS. If managers wish to view a dashboard that shows exactly how their resources are being used, the integrator engine will need to know (from the RIS) which studies are scheduled. Then, the integrator will need to receive status reports on each patient as the study is performed, and it must be able to show, at any moment, where the patient is located. The integrator engine will pull study information from the PACS indicating which specific imaging system was used, which technologist was responsible for the study, the time that the study was started, and the end of the study (indicated by verification). Unfortunately, in many institutions, the PACS and the RIS are not tightly coupled, so the integrator will need to match the information from both systems, synchronize the timing of events, and store the information for the analytics engine. Many methods have been suggested for doing exactly this job, and if an institution has the resources needed to create a custom application, those methods from the literature can be used. After the data have been successfully integrated into the business-intelligence database, an analytics engine can query the database and present the user with the information in myriad formats. If managers wish, for example, to see three CT units and their current throughput, they can be presented with a table showing the throughput over a user-selected time frame, the average length of time spent performing a study, and the queue of studies waiting to be performed. Managers can drill down to see where each patient is in the workflow and can highlight areas where the average waiting time or average imaging time is greater than expected. In addition to tables, business-intelligence dashboards and tools offer a number of other options for data visualization,14 including 2D graphs, 3D cubes, and tree maps.15 To illustrate a potential use of the tree map for radiology, Figure 2 is a fictional example of how a map might look for a department with three CT units, three ultrasound units, two MRI units, and six DR rooms.
image
Figure 2. Use of a tree map provides visualization of activity in a department with three CT systems, three ultrasound units, two MRI scanners, and six DR units. The sizes of the boxes represent the quantity of studies performed and the shades of the boxes represent the speed with which they are accomplished.
In each case, the size of a box indicates how many examinations are being performed on a unit (in comparison with other, similar units), and the grayscale indicates how quickly the examinations are being performed. Lighter grays indicate that studies are moving through the particular modality quickly, and darker grays indicate that studies are taking longer. The tree map is a constantly changing picture of the department’s throughput, with the blocks or leaves changing as workflow in the department changes. The black arrow is pointing to a DR room providing relatively fast throughput and also doing most of the DR examinations. The white arrow is pointing to the underperforming CT unit. In many cases, a tree map is interactive, and the user can click on a rectangle for more information—such as the location of the equipment, the personnel working with it, and the patients whose exams have been completed, along with numerical data on the average length of time needed for a study and the distribution of the performed studies. While it is tempting to require a tree map in a solution, multidimensional cube visualization of data, simple graphs, and tables could impart the same information. Radiology Reporting After studies have been completed, radiologists report their findings through dictation, usually using a speech-recognition system or a transcriptionist. The final step in the reporting cycle is the signing of the report by the radiologist, indicating that the report is complete and correct. Delays in the dictation and signing of reports can result in reduced turnaround time for a study (and this, in turn, can delay the patient’s treatment). In today’s more competitive environment, rapid turnaround can give a department an edge with referring physicians. A signal that a study has been completed is sent to a dictation system to alert the radiologist that a study is ready to be interpreted. The total turnaround time for the report is the time between completion of the study and the beginning of dictation, added to the duration of dictation and to the time elapsed between the end of dictation and the signing of the report. If a speech-dictation system is used, dictation duration and time between dictation and signature are automatically recorded as events occur. The completion of the study should be recorded in the RIS, triggered by an interface with the PACS. A business-intelligence dashboard can be created that shows each radiologist and his or her average total turnaround time, time from dictation to signing, and total volume of reports. In some cases, radiologists will not truly appreciate the power of these tools because they are understandably proud of their work; they could feel threatened by this type of analytics, which seems to value quantity over quality. Quality Metrics Khorasani10 has described some suggested measures of quality as performing the right procedure, using the right protocol, and imaging the right side of the right patient. To perform the right procedure, a referring physician might require assistance in an automated computerized provider order entry system that will guide the ordering of the appropriate procedure. Carefully documented, consistent imaging protocols will help ensure that the correct protocol is used to perform the procedure. It is the responsibility of the technologist to image the correct side of the right patient (after verifying the patient’s identity). Business-intelligence methods can be used to mine the data from these systems to see whether ordering physicians are ordering appropriate procedures and to ensure that the procedures were acquired using consistent protocols. Errors can be detected by examining the databases and reports for repeat procedures or for procedures performed on the wrong patients. Radiology departments have been challenged to reduce the radiation dose delivered to patients and to collect data for each patient showing cumulative dose, especially during CT and fluoroscopic examinations. At this time, many imaging modalities and PACS do not record and store this information, but systems are being developed that will extract these data from the modality or PACS and store the data in the RIS or HIS. Commercial Solutions If an institution does not have the in-house expertise needed to create business-intelligence analytics, and it does not wish to employ an outside entity to create a one-off solution, there are commercial solutions available to consider. The advantage of a commercial solution is that the vendor will probably continue to support and enhance the product, and it will continue to evolve over time. The disadvantage is that an institution will have to settle for a more generic solution, designed to meet the needs of a larger group of customers. If managers want to find the best commercial solution for their departments, they should identify the expected results and then issue a request for proposal to potential vendors. If the department and hospital have a single-source solution for the HIS, RIS, and PACS, that vendor will be the one most likely to provide the business-intelligence analytics needed by managers. Since vendors still have not implemented the types of systems described in the research-and-development literature, customers will have to put pressure on them to supply these systems. If enough radiology managers and radiologists demand business-intelligence solutions from their vendors, the vendors will respond. The RSNA® 2010 96th Scientific Assembly and Annual Meeting in Chicago, Illinois, and the annual meeting of the Society for Imaging Informatics in Medicine (to be held in Washington, DC, in June 2011) are both appropriate venues for starting the dialogue between vendors and their customers who need business-intelligence solutions. Janice Honeyman-Buck, PhD, FSIIM, is an imaging-informatics consultant and is editor-in-chief of Journal of Digital Imaging, the official journal of the Society for Imaging Informatics in Medicine; honeyman@medimg.com.