Extensible Stylesheet Language Formatting Objects (XSL-FO): a Tool to Transform Patient Data into Attractive Clinical Reports

Abstract

Clinicians at Wishard Hospital in Indianapolis print and carry clinical reports called “Pocket Rounds”. This paper describes a new process we developed to improve these clinical reports. The heart of our new process is a World Wide Web Consortium standard: Extensible Stylesheet Language Formatting Objects (XSL-FO). Using XSL-FO stylesheets we generated Portable Document Format (PDF) and PostScript reports with complex formatting: columns, tables, borders, shading, indents, dividing lines. We observed patterns of clinical report printing during a eight month study period on three Medicine wards. Usage statistics indicated that clinicians accepted the new system enthusiastically: 78% of 26,418 reports were printed using the new system. We surveyed 67 clinical users. Respondents gave the new reports a rating of 4.2 (on a 5 point scale); they gave the old reports a rating of 3.4. The primary complaint was that it took longer to print the new reports. We believe that XSL-FO is a promising way to transform text data into functional and attractive clinical reports – relatively easy to implement and modify.

INTRODUCTION

Despite hopes for a “paperless” world, paper continues to be a popular medium for delivering information, reading reports and viewing data. The growth of printer sales underlines this reality. One printer manufacturer, Hewlett-Packard, shipped more than 50 million printers in 2005, up 16% from 2003 – more than one per second.¹ The high resolution (1200 horizontal dots per inch or 7800 dots per width of a typical page – compared to 1024 pixels across the width of a typical screen) and strong contrast of print on paper provides an intrinsic readability advantage over video displays. Early studies reported faster reading and better comprehension when humans read paper compared to video screens^2,3; though this advantage waned in recent studies.^4,5

In health care settings, paper enjoys other advantages. Providers move rapidly between workstations. Their interactions with the computer are frequent, but short. The additional time to log on a computer and pick a patient record for viewing can be burdensome. Paper is the quintessential portable “device”. It is lightweight; it requires no login; and it does not run out of power. It supports easy annotation and quick navigation between pages; and it gives users the flexibility to arrange pages in physical space.⁶

Physicians who use the Regenstrief Medical Record System (RMRS) at Wishard Hospital print clinical reports called Pocket Rounds. Indeed, they say that it is one of the favorite features of the RMRS.^7,8 These reports display a patient’s Medication List, Allergies, Code Status, Problem List, Active Orders, Laboratory Studies, Radiology Interpretations, and more.

We have verified the popularity of these paper reports. In a study of PDAs at our institution, providers were instructed to use only the PDA to view patient data during the intervention period. Despite this requirement they continued to print daily Pocket Rounds reports at their baseline rate.⁸

From their inception in 1991, until February of 2005, we made no changes in the format or “looks” of the Pocket Rounds reports. At length we felt that Extensible Stylesheet Language Formatting Objects (XSL-FO) might offer opportunities for enhancing these popular reports.

The last few years have seen the emergence of XSL-FO – a World Wide Web Consortium (W3C) formatting standard.⁹ XSL-FO is a declarative language that can convert any XML tagged text into a sophisticated, publication quality report, using all of the graphical power available to modern-day printers. We hypothesized that with small changes we could convert our existing report generator to emit XML tagged data; that with modest effort we could use an XSL-FO processor to produce flexible publication quality reports; and that clinicians would prefer the new report format.

METHODS

We performed the study at Wishard Memorial Hospital: a 264-bed inner-city facility, affiliated with Indiana University School of Medicine, and described elsewhere.¹⁰ Clinicians at Wishard Hospital organize their workday around an order entry workstation called the “Gopher”. This system prints clinical reports on an 8½″ by 11″ sheet. Folded in half, it fits perfectly into a white coat pocket – hence the name “Pocket Rounds”.

For this study, we used XSL-FO to create a new version, which we called “Super Pocket Rounds”. (See Figure 2.) For this version, we embedded the patient-specific content in XML, and used XSL-FO table and list structures to organize the content. We enhanced the readability of the tables by converting from fixed space Courier to proportional space Times New Roman. We chose a serif font because of its greater legibility.¹¹ To make it easier to read the flow sheet, we added pale shading to alternate rows. We highlighted abnormal results and critical results with moderate shading and “reverse video” shading, respectively. We introduced a light line between columns to represent transitions from one day to the next, and a dark line to represent transitions between one month and the next.

Figure 2.

Typical snippet from the new version of “Pocket Rounds” reports.

Figure 3 illustrates the overall architecture of an XSL-FO report generating process. The XSL-FO processor accepts the XML content delivered by an application. First the XSL-FO processor transforms the XML data into formatting elements; then it renders these into a file format that a printer can understand, e.g., PDF or Postscript. Finally, the document server delivers this rendered document to a printer, either directly or through a print server.

Figure 3.

Architecture of the new printing process.

One must have an XSL-FO processor to make this work. A number of choices exist in the market, including an open source version from Apache. We used two different commercial XSL-FO processors in this project: Ibex PDF Creator 3.3 for .NET (Visual Programming Ltd., Wellington, NZ) and XEP 4.4 (RenderX, Inc., Palo Alto, CA). For this project we delivered the XML data stream from our Gopher system to an NT document server. This server ran a C# script that controlled the XSL-FO processor and forwarded the report to our Novell print server.

Users request the new reports from the same screen as the old reports. (See Figure 4.) The only difference is they now have a menu that gives them a choice of the old report format (still available) and three variants of the new report – one with 5 point, one with 6 point and one with 7 point fonts. (Figure 2 is an example of the 5 point font.) If the user chooses an old style report, the ASCII text streams directly to the Novell print queue and from there to the selected printer. If the user chooses a new style report, it is generated in the way described above.

Figure 4.

Screen shot of the Gopher user interface. Users can select three new versions of the reports.

The study occurred in two phases. At the beginning, our software (Ibex) could only render PDF reports. Because none of the older hospital printers supported Direct PDF Printing, we purchased three HP 4250N printers (@$1,177) and extra printer memory (Kingston 256 MB DDR module @$69), for the three busiest printing locations. Though these represented a small fraction of the hospital’s printers, they included 46% of the “Pocket Rounds” printing volume during the baseline period: enough usage to test our hypotheses. Eight months into the project, we mastered PostScript and found an engine that would render it (RenderX) – and were able to extend the project to ten more areas with HP 4000 and 4100 printers. With these additional ward locations we covered 90% of the Pocket Round printing.

The focus of the first and major phase of the study was the use of new XSL-FO based reports on three study wards that had printers capable of producing them. We collected Pocket Round usage data for two baseline months before the introduction of the new reports, and for eight study months afterwards. We added 10 more locations at the eighth month and ran the study for another four months with printing available in 13 areas. We logged usage data each time any report was requested: the user ID of the requestor, the date and time of the request, the format of the report, the destination printer, the workstation where the request was made, the number of patients included in the request, and more. We excluded all reports printed by investigators.

In the first months of the study, we asked users to complete a 22 question survey instrument. We asked them to rate their satisfaction with the old reports, and the new reports, using a 5-point Likert scale. We also asked specific questions about the changes we had introduced, whether they were troubled by the printing speed, and whether they could find results faster in the new reports. This survey was given during morning report and lunchtime pizza sessions. One of the investigators (LS) used a stopwatch to measure printing times under various configurations. Measurements began when a user entered a request, and ended when the last page came out of the printer.

This study was approved by the Indiana University Institutional Review Board.

RESULTS

During the eight month study period, 772 clinicians requested 26,418 Pocket Rounds reports on the three study wards. Of these, 40% were medical students, 38% medical housestaff, 8% attending physicians, 4% nurses, 3% pharmacists/pharmacy students. Attending physicians did not request reports (either new or old reports) as often as the student and resident members of the ward teams. However, attending physicians constitute one of eight members of the ward team, so the numbers should have been less. They tended to request larger numbers of patients at once. (See Figure 5.) The distribution of users was similar to that of 1114 users who requested 43,525 Pocket Rounds reports from all workstations (including those without the “super” report option). One exception was pharmacists, who made up almost 8% of users throughout the entire hospital, but had limited access to the study printers.

Figure 5.

Number of Patients per Report, by User Role.

The uptake of these new reports was rapid. Within a few days of their introduction, requests for the new “Super Pocket Rounds” grew to more than half of all requests on the study wards. We made only a modest promotion effort: three brief announcements at morning reports; three at pizza lunches. Of the 26,418 reports generated during the eight month study period, 20,583 (78%) were printed using the new system. At the end of the study, this rate of usage was approaching 85%. (See Figure 6.)

Figure 6.

Monthly counts of reports printed on the three study wards. Comparison of new “Super Pocket Rounds” reports with old “Pocket Rounds” reports. Shown are two baseline months and eight study months.

During the first eight months, providers working on non-study wards (those that could not print the Super Pocket Rounds) often shipped their requests to one of the 3 study wards. The new report was worth the price of a hike to another ward. Indeed, requests from users on other wards accounted for 16% of printing on the study wards. Such requests accounted for only 8% of printing during the baseline period.

We gave users three choices of font size for the new pocket rounds (something easy to do with XSL-FO). In order to keep the pocket rounds report small we included only one row of data per variable; so the price for the larger font was fewer columns of data. Half of the time (53%) users requested the smallest (5 point) font and the most compact (11 columns) report. The 6 point (9 columns) and 7 point (8 columns) reports were less popular: 24% and 22% of requests, respectively.

The Gopher lets users request a report for one patient at a time, or for a set of patients. The majority of report requests (58%, or 15,358) were for a single patient’s data. However, this number underestimates the total reporting volume because many requests were for a set of patients (e.g., all of the patients on the physician’s team). The average number of patients per report was 4.2: over the eight months of the study, 26,418 reports represented requests for 112,015 individual patient summaries.

Completed surveys were obtained from a convenience sample of 67 clinical users: 30 medical residents, 27 medical students, 2 attending physicians, and 8 pharmacists/pharmacy students. The respondents represented roughly two thirds of those approached. Results of the survey instrument confirmed a preference for the new XSL-FO formatting. Satisfaction was rated on a five point scale as 4.2 for the new reports and 3.4 for the old. The difference was significant (p<0.001) using a paired t-test. Thirty two users (48%) agreed that the new reports were “much better” than the old reports; an additional 19 users (28%) said they were “slightly better”.

Most respondents thought that the highlighting of abnormals (94% of users) and panic abnormals (87%) with different shading was useful. Most (79%) thought it was “faster” to find abnormal values in the new reports than in the old. The light shading of every other row was favored by 78% of users; the thin line that demarcated columns from different days was favored by 84% of users. Their stated font size preference closely echoed the usage distribution described above.

The new reports took longer to generate and print than the old ones. For example, a typical 13-patient report (20 pages) required a mean of 130 seconds (SD 13 seconds) to complete printing in the new format. The same report was produced in 55 seconds (SD 3 seconds) in the old format. Users noticed the difference. More than half of survey respondents (55%) said the new reports took too long to print. However, within this subgroup nearly half (43%) were not bothered by the slower printing. Printing time averaged to 10 seconds per patient for multiple-patient reports.

Verbal feedback was obtained from housestaff and students during pizza lunch discussions and during morning report. Overall, the new reports were received with enthusiasm. The predominant complaint was the increased printing time. A few users informed us that they continued to use the old reports for this reason.

When we were able to master the usage of PostScript we extended the Super Rounds reports to 10 more locations (90% of all of the printing volume). We did not carry out a promotion of any kind for this next phase. Yet there was a fast uptake at these locations also: to 50% usage of the new reports within six weeks and to 65% after four months. Combining the data from all 13 areas (with Super Pocket Rounds capable printers), during these four additional months 13,408 (71%) reports were printed using the new system.

DISCUSSION

Through their actions and their survey responses, users indicated a substantial preference for the new XSL-FO formatted Pocket Rounds reports. They voted with their choices despite the fact that printing the new reports took longer.

The typography literature indicates that choice of fonts and layout makes a difference in the speed and accuracy of perception.^11,12 We suspect that the preference of our users derives from the perceptual utility – and not just the pleasant appearance – of these reports. That is, they could digest the contents of these new reports faster than the old – but we did not obtain data to verify this impression. We plan to experiment with mixes of graphic displays for frequent measurements and to study the speed and accuracy of abnormal detection.

Other researchers have investigated the importance of the format in which data is presented to the clinician. Fries established that a medical record with fixed-format, flow sheet organization permitted physicians to access data four times faster – and with greater accuracy – than other print formats.¹³ Changes in the display of microbiology results – albeit on a computer monitor – improved access times and error rates.¹⁴ Another study of printed displays of lipid tests documented faster physician retrieval.¹⁵

XSL-FO made it easy to develop sophisticated reports, to modify them, and to offer rich user tailoring such as changing font sizes and ordering lists. To users XSL-FO takes care of all of the table layout work, the wrapping of columns, and pagination; and it can accommodate photographs, watermarks, and graphics. And it is relatively straightforward to insert XML tags to replace the print layout control in the legacy report programs.

XSL-FO made it easy to give users a choice of font size. In our study, most clinicians selected the smallest (5 point) font size. The larger fonts (especially in the 10 to 12 point range) are known to be more legible.¹² However, a small font size allowed 11 data columns to squeeze into each table. Apparently clinicians valued the dense packing of information in these reports – perhaps because it minimizes the need for page turning.

We learned a few lessons. In principle the XSL-FO specification supports many print formats including PDF, PostScript, Rich Text Format and HP PCL. But currently many XSL-FO processors support only PDF. PDF output works well but only if one has a printer that supports Direct PDF Printing – which means a new and medium priced printer, and possibly the installation of extra memory. In one case we could not get direct printing to work on a new printer that claimed to support it. It is better to get a formatting engine that supports PostScript Level II, which is available on most printers and works well. The printing speed of PostScript formatting was also slightly faster than the corresponding PDF.

We are enthusiastic about XSL-FO. It provides a flexible mechanism for generating clinical reports – easy to implement with only modest changes to existing software. We are implementing a similar reporting system at a second Indianapolis hospital, which uses a different data retrieval application.

Furthermore, these patient summaries constitute an infrastructure for more advanced presentation of information. Segments of EKGs or photographs could be introduced easily, as could decision support reminders: a few additional lines of programming could trigger a suggestion for flu vaccination based on a patient’s age. We know that such reminders would have a broad reach, because hundreds of hospital clinicians are generating and reading these reports.

Figure 1.

Typical snippet from the old version of “Pocket Rounds” reports.