Radiology information systems: evaluation and selection issues.
Even the way people use computers has changed dramatically. Today, it is common for business computer users to interconnect through networks. In addition, the Internet gives users almost instantaneous access to a vast array of information, and electronic mail is fundamentally changing the way computer users communicate.
Computers have revolutionized modern radiology departments as well, although it can be argued that the computerization of radiology administration lags behind the computer revolution in general. Recent computer advances should benefit radiology departments through lower prices, enhanced productivity and flexibility not possible with older hardware platforms.
A radiology information system (RIS) is a computer system designed to perform administrative chores and provide accurate information in the radiology department or clinic. Typically, an RIS captures vital business and patient care data and provides functions such as billing, results reporting and film folder check-out. The manner in which systems accomplish these tasks varies greatly.
Users of radiology information systems have a right to be demanding of vendors on issues related to cost, productivity, flexibility and communication between computer systems found in a radiology department. In addition to system features, buyers should consider ease of implementation because poorly implemented systems make users work for the computer instead of the computer working for its users. This article is designed to help radiology department managers evaluate and select an RIS.
Equipment Platform Issues
A radiology information system should operate on the simplest, most economical and modern hardware platform that will provide acceptable performance. Buyers should be aware of the advantages of radiology systems running on microcomputer network servers or client-server database servers compared to the older terminal-based technology. Modern platforms offer better economy and more flexibility.
One reason for this is the accelerating performance gains in microcomputer and network technology in the last 15 years. These gains are achieved almost continuously with respect to previous microcomputer products. Gains are even more pronounced when compared to bigger, centralized minicomputers and mainframe systems, which have longer product cycles. (A mainframe is a large computer that may have many attached terminals; minicomputers are intermediate between mainframes and microcomputers in terms of size, power and storage capacity.) Serial communication used by older, centralized systems is orders of magnitude slower than current network communication protocols such as Ethernet.
With modern distributed processing architecture, more work is performed locally at the operator's workstation rather than centrally at the mini- or mainframe computer. Networked workstations communicate with file servers only for disk and printer access. Client-server based systems that use a centralized database server (usually a minicomputer) solely for database functions are quickly gaining popularity. All other processing is done on local microcomputer workstations. The key benefits of modern platform technologies are distribution of processing, much faster communication and greatly reduced costs. For the end user, this means improvements in economy, flexibility and performance.
Cost Issues Related to Hardware Acquisition, Maintenance and Upgrades
Because of the economics of large-volume manufacturing and sales, microcomputer systems and networks are great values, especially compared to the small, proprietary mini- and mainframe market. In addition, microcomputer performance is constantly improving. This means that the same cash outlay buys ever-greater capabilities such as increased processor speed, larger and faster disk drives and more capable printers.
Minicomputers are still expensive to buy and operate. Many require expensive proprietary operating systems and licensing fees. Maintenance on these platforms can be expensive, often requiring costly service agreements with the manufacturer, who may be the only service vendor. Microcomputer repair, on the other hand, seldom runs more than a few hundred dollars and can be obtained from local vendors, often at the customer's site. These systems are inexpensive enough to keep a spare on hand.
Upgrading a minicomputer also can be expensive. Upgrades generally consist of proprietary hardware or operating systems available only from the manufacturer. Upgrades are offered infrequently compared to the microcomputer market, where faster and more advanced products are announced daily. Ongoing technological advances and lower prices are available immediately to users of microcomputer-based platforms.
Before the graphical user interface (GUI), minicomputers generally required manual typing for input fields where a selection was made from a fixed list of entries. Better software allowed the use of numbers, shortcut abbreviations, mnemonics or bar code readers. The benchmark of data entry for RIS software today should be the level of user-friendliness in current commercial microcomputer products. Buyers should no longer accept manual typing as the sole method of selecting from a fixed list of entries. Graphical interface tools such as choice lists and check boxes are simpler and reduce errors while speeding up data entry. One particularly effective input scheme, incremental typing, is found in the help index of many commercial microcomputer packages. Based upon successive keystrokes, the computer displays the most likely entry. The user accepts an entry by pressing the space bar or the tab or enter key.
Any time an RIS can reduce keystrokes, it improves user productivity. One effective way to do this is to provide appropriate default responses for fields where a response can be predicted. The default response is the one selected if the user takes no action in a particular input field. One way to implement the process is to allow designation of a default response for an individual field. This feature is common in commercial microcomputer database software. An operator always should be able to override a default entry by simply typing in a new response.
Some software allows a field's last entry to serve as the default response or automatically provides a choice list from recent entries. A good example is the file menu of many commercial microcomputer programs. At the bottom of this menu, the user can select from the last several files accessed. Regardless of the technique used, the goal is to reduce keystrokes by having the computer select a response for the operator that is likely to be correct.
RIS vendors try to develop a standard product that will work for all customers. Although this makes life easier for them, buyers should make certain that such standardization does not negatively impact productivity. If a small fraction of a vendor's customers require a particular feature or field, a well-written RIS should not allow that feature to inconvenience or slow down other users. Systems should be designed to allow users to designate fields where an entry is mandatory and automatically skip fields that the facility does not use. While one tab keystroke does not take long, these seemingly small items can affect productivity because screens related to registration, tracking or billing may be used thousands of times a day in a busy department. In today's climate of shrinking budgets and staffs, administrators cannot afford inefficiency. Graphical interface tools eliminate the need for users to respond to sequential prompts. At a minimum, RIS data entry screens must allow users to move easily between input fields and go back to correct errors. A well-written user interface does not allow a user to place the cursor anywhere but in designated input fields.
The ability to customize input fields by masking and checking data for validity should be mandatory in an RIS. Data masking allows fields to be customized so that only keystrokes consistent with the field's data type are accepted. For example, if a hospital uses numeric patient identification, a properly masked input field will accept only numeric key entries while ignoring any other character keys pressed. Not only does this prevent errors, it eliminates confusion when the upper case letter "O" or the lowercase letter "l" are used for the numerals "0" and "1."
Field masking also should provide separators such as the slashes in dates or the hyphens in Social Security numbers and telephone numbers. Separators should be visible in the field and the cursor should jump over them automatically. This reduces keystrokes and prevents potential errors.
Field masking often is combined with validity checking. Software constrains data to meet previously defined criteria for a particular field; otherwise, the entry is not accepted and the cursor will not move from the field. For example, if a hospital uses eight-digit patient numbers, an RIS capable of validity checking will only accept entries that are eight digits long. Field validation also will refuse to accept obvious errors, such as the date "02/31/96." Date fields may be constrained in certain uses such as date of birth, prohibiting entry of dates in the future. Of course all current software should be designed to correctly handle date calculations at the turn of the century. Validation also may be used to constrain entries to an acceptable range of values. For example, negative or noninteger numbers should not be an acceptable entry for film count.
Software vendors have been known to take short-cuts to accommodate a large customer base with varied ways of doing business. One system offered a blank 20-character line for the patient identification field and accepted keystrokes of any kind. For an institution that uses an eight-digit patient identification number, this is an example of achieving uniformity at the expense of proper implementation. Masking and validity checking are standard features of microcomputer database software, and buyers should expect no less from an RIS.
Another way an RIS can help prevent entry errors is by verifying permissible but suspect data. For example, it should not assume that a patient ID number manually entered by an operator during registration is correct. Instead, it should require the operator to enter the patient ID number twice. Also, it is helpful when adding a patient if the RIS notifies the operator when it finds patients with similar or identical demographics (eg, name, race, sex and date of birth) under a different patient ID number. Such features can prevent a simple transposition of numbers from turning into a duplicate patient entry. While requiring a little more time up front, it is better to find errors during data entry.
The presentation of output data is another area where an RIS should be held to commercial microcomputer software standards. Proprietary data viewing and analysis tools should not be the sole mechanisms for administrators to access data. Customers must be allowed to analyze data in ways not anticipated by RIS developers. Buyers also should look for an open architecture that makes all or user-selectable RIS data available for on-line analysis or for export and off-line analysis.
ASCII text files with fixed field lengths or special delimiting characters can be imported into most microcomputer analysis software. It also is acceptable to provide data in the format of industry-standard spreadsheet or database programs such as "dbf" (dBASE), "mdb" (Access), "xls" (Excel) or "wk3" (Lotus). Such standards can be handled by virtually all commercial products and are readily translated. The RIS should provide such data with no programming or special hardware adaptation required.
The literature is full of examples of costly and time-consuming efforts by users to access their data on proprietary platforms.[1-6] These efforts generally involve special hardware rigging such as passing minicomputer terminal or print data through a custom-programmed microcomputer. Although such efforts indicate superior inventiveness, buyers should try to avoid them. In each case, once information was made available in a microcomputer environment with its wealth of tools, custom analysis became easy.
A modern RIS should allow distribution of clinical patient result reports by fax machine, e-mail and network printing resources, in addition to on-site printing. The RIS also should provide flexibility in sorting printed reports. Some departments with an RIS still print and distribute multiple paper copies of exam reports, such as a copy for the film jacket, one for medical records and one for the referring physician. The RIS should be capable of batch printing each type of copy separately and in any desired order. For example, the RIS should allow the user to print the film jacket copy in the order films are filed, such as in terminal digit order. Physician copies may require sorting by physician, practice group or building. A well-implemented report distribution system should completely eliminate hand sorting of computer-generated reports.
Radiology systems must communicate with a number of other computer systems in order to effectively share information. Stand-alone radiology systems -- those that are not an integral component of a larger hospital system -- generally include a connection to the hospital to accomplish functions such as results reporting, billing and admission/discharge notification. Such interfaces are a great help in automatic data verification, entry of patient demographics and ensuring that reports are delivered to each inpatient's current location. Because such interfaces are extremely useful in terms of reducing data entry and eliminating steps where errors can occur, nothing less than their full implementation should be accepted.
There are a number of ancillary or special purchase computer systems for radiology. Examples include digital dictation, speech recognition, mammography and PACS. While each vendor emphasizes various special features of their systems, data entry is sometimes addressed as an afterthought rather than a primary consideration. Much of the data required by these ancillary systems already exists on the RIS. Redundant data entry is unproductive and its elimination should be a primary goal of buyers. Even if the amount of data to be entered is small, the frequency of this redundant activity affects productivity. The goal should be to automate all routine data entry operations to improve both accuracy and efficiency.
Interfacing should be a major consideration when purchasing an RIS or an ancillary computer system. RIS architecture should be open, providing data "hooks" for ancillary systems. Connectivity to the RIS should be a primary consideration in the selection of ancillary systems. One effective way for a buyer to cope with interface issues is to write specifications assigning vendors of ancillary computer systems total responsibility for interfacing to an existing RIS. Responsibility does not dictate which company will actually develop and install the interface. The goal of a properly written specification is to ensure that the facility does not end up without an interface while two vendors point fingers at each other. Specifications for ancillary systems should indicate that the completion of a working interface is an integral part of the product and will be required before an order will be signed as complete and the vendor paid.
Potential RIS buyers should be satisfied that a system is offered on the most economical platform that will support the product. Users also must take care that the RIS is working for them by providing maximum flexibility while minimizing data entry time and errors. Microcomputer systems are not only less expensive to buy, maintain, support and upgrade, they also give the user a level of flexibility impossible with larger centralized systems.
Users often accept less than they should in RIS software. Vendors should not dictate what buyers need. The vendor's role is to translate the buyer's need into computer code, not to define the buyer's problems in terms of their isolated environment. Users should evaluate vendors in part on their responsiveness to productivity issues. A vendor who is more worried about product uniformity between customers than about a customer's productivity is cause for concern.
When evaluating RIS software, buyers must look beyond features and concentrate on implementation. Modern input design minimizes errors and reduces data entry time.
In addition, buyers should carefully evaluate interfaces between various radiology computer systems. Vendors must be pressed to design products that can be readily interconnected, and they should accept responsibility for implementing such interfaces.
[1.] Donnelly LF, Johnson ND, Taylor CN. Increased efficiency of radiology information management with a radiology support system. AJR. 1997;168:611-612.
[2.] Frank MS, Berge RE. Value of on-line informational databases in a radiology department. AJR. 1995;164:1537-1539.
[3.] Frank CL, Stewart BY, Rowberg AH. Use of data in a radiology information system for labeling computed radiographs: an interface to connect the two systems. AJR. 1995;164:745-747.
[4.] Crabbe JP, Frank CL, Nye WW. Improving report turnaround time: an integrated method using data from a radiology management system. AJR. 1994;163:1503-1507.
[5.] David G, Gregory S. Adding a microcomputer barcode network to a minicomputer-based radiology information system. Journal of Digital Imaging. 1990;3(1):31-33.
[6.] David G, Gregory S. Mini-micro-mainframe computer marriage: combining technologies in a radiology results reporting system. Journal of Digital Imaging. 1989;2 (1):27-30.
George David, M.S., is a physicist in the Department of Radiology at the Medical College of Georgia, Augusta, Ga. Over a period of 12 years, he developed and maintained the department's RIS system. He also has been involved in the purchase of two commercial radiology information systems as well as several ancillary computer systems.
Reprint requests may be sent to the American Society of Radiologic Technologists, Communications Department, 15000 Central Ave. SE, Albuquerque, NM 87123-3917.
[C] 1998 by the American Society of Radiologic Technologists.
|Printer friendly Cite/link Email Feedback|
|Date:||May 1, 1998|
|Previous Article:||Radiology of Chest Diseases, 2d ed.|
|Next Article:||Cholangiopancreatography using magnetic resonance.|