Almost 25 years ago I co-developed a prototype of a radiology imaging workflow system at Shadyside Hospital in Pittsburgh. We published details in the 18th Annual Symposia on Computer Applications in Medical Care (SCAMC, now AMIA). I’ve been meaning to resurrect some of that detail for a long while. I finally have.
We’d been tasked with replacing a radiology information system from the sixties with a system based on what was then current tech. By the way, if you follow along the screens, you’ll see it is indeed a workflow. The next step would have been to control tasks and screens from a workflow engine, though we didn’t get that far. Anyway, this was when my interest in healthcare workflow management systems technology really gelled. SGML? Standard Generalized Markup Language, which predates HTML. This system was essentially a clinical domain-specialized web-browser (and back-end servers) before the advent of the World Wide Web.
FELIX (FELIX Enables Limitless Information Exploration) provides a generic graphical interface for browsing medical and administrative information. FELIX relies on a large number of industry standards such as TCP/IP, X11R5, SQL and SQL-II, Postscript and SGML. In particular, the interface was developed using a high-level X-windows graphical scripting language called Tcl/Tk (Ousterhout, 1990, 1991,1994). We used real patient information and based our targeted applications on analysis of patient care processes at a local urban community referral hospital. We intended FELIX to be a vision of what is possible, a prototype to force us to confront the necessary integration of disparate technologies, and an inducement for clinicians and administrators to press for open systems file formats, programmatic interfaces and network protocols.
A magnetic resonance image scan has been completed and is ready for interpretation.
By pushing the button titled “Record Report”, a tape recorder with the standard buttons appears on the screen, which allows the physician to verbally enter their findings.
A document that looks like a “document” (left), but which is actually derived from a database, can be printed, FAXed, or electronically mailed.
Patients are tracked into and out of rooms using bar code readers on a network and the information is periodically used to update bar charts. Each bar represents a room. Height represents length-of-stay. Red indicates length-of-stay larger than an adjustable threshold.
The digitized image of a peripheral blood smear has annotated areas that correspond to high-lighted phrases in the textual clinical report.
This resource scheduler represents doctors, nurses, rooms, and equipment across the top, and times down the side. Constraint processing indicates in green the times during which resources are available to be committed together.
On the left is a list of procedures and referring physicians, which can be used to plot procedure volume over time. Decreasing rates of referral may suggest need to intervention.