Natural Mapping and Self-Evidence
System A’s successful use of natural mapping and self-evidence () makes its insulin orders the easiest of the three to navigate. The design reminds the user of a familiar paper order form (). There is a tabular format with columns for the time of day and rows for type of insulin. This format matches clinicians’ mental image of insulin ordering. Insulin orders in the CPRS () and system B () instead appear in the same format as any other order in those systems, failing to reflect the uniqueness of insulin ordering.
Definition of Terms: Usability Principles
Commercial system A, tab A. Basal and nutritional insulin
Commercial system B. Prandial dosing.
Tabs are another important and effective design feature. Krug9
believes that tabs are excellent because they are “self-evident,” “hard to miss,” and “suggest a physical space.” Tabs are a rare example of a physical metaphor on a computer screen. Both system A and system B use tabs well, making the active tab’s color bleed into the page below, showing a physical connection between the two, distinct from the tabs in the background.
System B uses another effective design convention, with small arrows that point sideways to tell the user that there is a submenu behind that selection (). When opened, this arrow changes to a downward arrow. This is a common convention that requires no mental effort for a user to decipher.
Commercial system B. Low dose regimen.
One place where system A shows ineffective design strategy is on its tab B (), labeled “Meal Time Correctional Insulin.” Here, the headings at the top, written in blue font, are not clearly connected to the table below, in black font. While font color was used to denote a difference, the designers also could have used vertical lines effectively so that the user could immediately, without thinking, connect the insulin units to the blue headings above. Another problem in tab B is that the type of correctional insulin chosen is not clear to the user. While it is in fact the same type of insulin as chosen in tab A, the user might have to do a few clicks back and forth between tabs A and B to discover this fact.
Commercial system A, tab B. Mealtime correctional insulin.
The three systems use feedback to varying degrees. Effective use of feedback is critical to let the user know that his action has ended in a result or made a change in the system. In the CPRS, feedback is used when buttons turn blue and are underlined after they are clicked.
System A lacks effective feedback to the user in that the user can click “save and exit” after completing tab A without ever looking at tabs B and C (). There are prechecked orders on tab C, meaning that the user is signing orders without ever even seeing them or knowing they exist. The system should have some feedback to let the user know that he has unwittingly chosen these orders. The system might instead have chosen to force the user to move to the “next tab” until he has seen all of the tabs and can then “save and exit.”
System B effectively uses feedback by displaying the chosen dosage of insulin on the screen, 2 units in this example, immediately after the user enters the information. However, it then displays both the breakfast and lunch doses of insulin as “starting today at 1656,” giving no indication or feedback to the user that these are in fact two distinct orders and not erroneous duplicates ().
Both system A and system B effectively use affordance. In system A, fields where an entry is required are shaded yellow (). In system B, required fields similarly use affordance, displaying a red stop sign with an exclamation mark ().
Constraints are used frequently by these systems both effectively and ineffectively. Norman18
writes about constraints and how they are used to limit the user’s choices to make it impossible to do the wrong thing. Alternatively, a constraint can be designed to signal the appropriate action. One type of constraint used by CPOE systems are defaulted orders. This would have been useful in system B’s “Notify Physician Parameters,” where it is likely that these three orders are chosen by nearly every clinician using the order set (
). Prechecking these three boxes could have cut down on mistakes as well as save time.
System A makes more effective use of constraints for the same types of orders, going even further by removing the checkbox next to these orders, making it impossible to deselect them (). The designers, knowing that these orders should be chosen every time, did not want to allow the user to deselect them. This intentional choice saves the user the cognitive workload of having to pause for even a moment to consider whether these orders should be selected. In tab B, system A has pre-selected the sensitive correctional insulin scale based on the patient’s body mass index (). This is an example of more advanced decision support, potentially saving the user time, standardizing care, and preventing user error. However, this feature could also introduce error, for example, if the body mass index was incorrectly calculated based on an incorrect height entered into the system.
Another type of constraint is a “corollary order.” With “corollary orders,” selecting one option automatically opens up more orders that the system deems linked to the first one. In the CPRS, when an “acute care insulin sliding scale” order set is selected, the user receives an automatically progressing list of prechecked orders (). This list includes the insulin orders, PRN glucagon orders, and “call house officer” parameters. Once the preselected list is opened and the user begins navigating through, the user is actually unable to do anything else until each order in the entire set is either confirmed or rejected. This design is very effective for standardization, guiding users to select the exact same orders every time with very little thought necessary. However, in cases where customization is desired, there is no signal to the user or prompt for this to occur. Of note, this “automatic” order set in the CPRS does not include orders for finger stick glucose checks. It seems rather unlikely that a user accustomed to quickly clicking through the order set is going to realize that a separate order needs to be placed for finger stick glucose checks.
The dosing schedule choice in the CPRS shows both constraints and a lack of constraints. While insulin is nearly always given in one of 3 or 4 dosing schedules, the system presents a scrolling list of over 50 dosing schedules, the same list presented for any medication. Shortening this list could both save time by diminishing the cognitive workload of the user and also reduce the potential for error. Error could occur here either by a provider accidentally choosing the wrong dosing schedule or by mistakenly believing that all these options are realistic and correct choices. Koppel and colleagues6
surveyed house staff and found that one common error caused by CPOE was the mistaken belief that, if an option is present, it must be viable. The CPRS does try to avoid errors by preselecting the “QID AC+HS” dosing schedule, which is likely to be the most frequent selection.
A potentially dangerous lack of constraint exists within the CPRS. In the dialog box where a regular insulin sliding scale order is written out, there is a free-text section where a provider can change the insulin doses that correspond to each blood glucose range. Since this is a free-text section, there is no limit to what number can be placed here and thus no limit to how much insulin could be ordered. One of the primary advantages of CPOE is dose-range checking, which is absent here.
In system A’s tab A, a user could potentially choose multiple types of short-acting insulin. A constraint could have been used here to allow only one type of short-acting insulin to be chosen (). The designers chose not to do this, allowing for the rare clinical event where a provider might want to order multiple types of short-acting insulin. This raises a question that challenges many areas of CPOE beyond insulin ordering. Should constraints be designed in the system to prevent a provider from doing something that is usually improper but rarely might be correct?
System B takes a different approach from system A and the CPRS when it comes to preselecting correctional insulin doses. Though the user can open a submenu for low, medium, and high correctional doses () and there is an instructional guide at the top once the submenu is opened, the user must still click on each individual insulin order. This process of having to make multiple individual clicks rather than a single click to place a group of related orders may introduce a relatively high risk of error.
While both system A and system B clearly remind the user as to which patient’s chart they are placing orders in, the CPRS lacks this critical safety feature. Potentially adding to a user’s frustration would be the possibility that the user would have to click through the entire “automatic” order set only to realize that the work was completed in the wrong patient’s chart, and now has to start over from the beginning.