Abstract
Limited prospective data exist that define advantages and disadvantages of text-based versus graphical user interfaces in Computerized Provider Order Entry (CPOE). We created a single set of admission orders in both formats and, using a randomized crossover design, assessed the usage and perceptions of the two interfaces by 51 physicians during the first six weeks (2 week blocks of one format, then the alternative, then free choice) following implementation in a postsurgical unit.
Methodology
Care following surgery is provided by 51 attending anesthesiologists at our institution. The commercially available CPOE system implemented at our institution allows for both text-based input as well as input via a graphical user interface. We elected to implement both text-based “orderset” (OS) and graphical “iForm” (IF) formats for entry of the same set of admission orders for the first 6 weeks of use of CPOE on our post-anesthesia care unit. The anesthesiologists were randomly assigned to one of two groups, A and B, and those arbitrary designations were also used to label the two ordering formats to direct the anesthesiologists to the appropriate format for a period of two weeks each (e.g., the group A users were directed always to select Format A orders, which pointed to the OS format for the first two weeks and to the IF format for the second two weeks). Four weeks of directed choice were followed by two weeks of free choice. Data was obtained from the database of orders placed, the usage logs, and two surveys completed by the users.
Evaluation Results
Users predominately chose to use the graphical (IF) format: it was used for 70% of the orders in the free-choice phase, and 17 of 24 (71%) of survey respondents preferred the graphical format. The text-based (OS) format gained substantial support, however: if keeping both formats were an option, 15 of 26 (58%) answered that they would choose to keep either both formats or the text-based alone; and those users initially assigned to the graphical format were actually somewhat more likely than their counterparts (36% vs. 21%) to prefer the text-based format. Experience level (based on the number of orders placed) had a small but significant (p = 0.02) correlation with preference of format, with more experienced users preferring for the text-based format. According to time measurements from the usage logs, CPOE sessions in which the graphical format was used averaged 27 seconds shorter (162 vs. 189 seconds, p <0.01). No statistically-significant differences between graphical and text-based formats were found for length of stay, rate of mistakes made, or the number of orders for diagnostic tests or medications; miscellaneous orders were placed slightly more frequently (5.44 vs. 5.14 orders per session, p = 0.03) from the text-based format. Development time was higher (40 vs. 28 hrs) for the graphical format. In surveys, users rated the graphical format superior for time required to use, time required to achieve comfort, ease of use, satisfaction, appearance, speed, and suitability for busy times of the day; the text-based format was rated superior for flexibility and suitability for patients with more acute or chronic illness.
Conclusions
The graphical format of CPOE orders in this commercial system was, on the whole, both more highly rated and faster, and it should be chosen if initial user satisfaction and speed are the primary goals and development resources are available. The text-based format, however, did gain a substantial minority following, was rated as advantageous in some specific scenarios, required less development time, and may show an increasing preference by users as they gain experience. Substantial efforts to tailor the capabilities of a system to requirements are required in either interface environment. Neither interface is so obviously superior as to exclude the other, and development efforts should be directed toward enhancing each to incorporate the strengths of the other.