Planning Your Implementation/Intervention: Designing Reliability and Other Methods to Ensure Successful Implementation

Greg Maynard and Jason Stein

Your team may come up with a dozen interventions to optimize prevention of hypoglycemia and improve glycemic control for patients in your medical center. Key interventions every team should implement are standardized insulin order sets and protocols for critical care settings, non-critical care settings, transitions, and perioperative situations. In reality, you are creating a series of linked protocols.

Why a Protocol?

The key concept is routine. Doing a complex activity the same way each time is the best way to make sure nothing is left out. In the hospital, protocols serve that purpose. They standardize and structure care delivered by a group of providers.

Why is routine important? Across a population of patients, one of the most common sources of suboptimal care arises from provider inconsistency. For a variety of reasons providers inevitably vary care inappropriately, whether compared to each other or compared to themselves. In fact, a graph that depicts improved system performance over time almost always shows a progressive narrowing of the range of performance data points. In a powerful way, protocols improve care by specifically reducing this unnecessary variation in performance, from medical decision making to ordering.

The best protocols, however, preserve our ability to customize care for special patient situations or circumstances. In contrast to variation that arises from provider behavior, variation from the protocol because of special patient situations is always acceptable. The protocol should make that clear.

As always, the devil is in the details. An order set/protocol will usually fail unless the team pays attention to these details.

A review of some principles for effective implementation is in order:

Principle 1
Keep it simple for the end user. There will inevitably be trade-offs between the depth of detail of guidance you want to give providers and the simplicity of the forms and the processes the end users have to go through. Most of the time, simpler is better. Minimize calculations the end user has to make or automate that process for them.

Principle 2
You can't interrupt the work flow. Don't become shortsighted about the importance of this particular intervention to enhance glycemic control and insulin administration in your medical center. Remember that this is rarely the primary focus of members of a caregiving team, and they are likely to be attending to dozens of other tasks per patient. Involve frontline workers to make sure your plans are feasible and that your order sets/protocols are easy to use. Check boxes and prewritten scales can encourage rapid acceptance because they make the work easier. Get their input on how to make implementation go smoothly. Clinicians should want to use your order sets if they are constructed properly.

Principle 3
Design reliability into the process. Human beings are incapable of doing anything reliably 100% of the time in the complicated health care setting. Part of your team's job is to engineer higher reliability into the process of getting rational insulin dosages to hyperglycemic patients. If your protocol relies solely on these traditional methods, you will be disappointed with the results:

• Common equipment and standardized order sets;
• Personal checklists;
• Working harder next time;
• Feedback of information on compliance;
• Awareness and training.

All these methods are helpful (and some are necessary), but they are not sufficient for achieving breakthrough improvement. You must design at least one of the following methods into your interventions to enhance the probability that each patient will receive the correct kind of therapy for his or her particular situation.

High-Reliability Strategies

• Desired action is the default action (not doing the desired action requires opting out).
• Desired action is prompted by a reminder or a decision aid.
• Desired actions are standardized into a process (take advantage of work habits or patterns of behavior so that deviation feels weird).
• Desired action is scheduled to occur at known intervals.
• Responsibilities for desired action are redundant.
• Algorithms and reminders are incorporated into the order sets.

Examples of these methods as they apply to insulin administration:

• Incorporate glycemic targets and HbA1C orders into the subcutaneous insulin order set.
• Integrate guidance on appropriate glucose monitoring and insulin dosing into your order sets.
• Integrate guidance on what changes to make in the insulin regimen when caloric intake is interrupted.
• Make subcutaneous insulin regimens with scheduled basal insulin the default.
• Strip out all other insulin order sets of your institution. A review of post-op, transfer, and admission order sets that all services use will probably reveal a half-dozen or more embedded sliding-scale insulin order sets that need to be eliminated. This will help to make your order set the only easy way to order insulin. Of course, correction insulin alone may be appropriate for a minority of patients, so your orders need to preserve this as an option while still encouraging basal insulin for most patients.

Examples of these and much more are presented in later sections, along with specific tips on how to integrate high-reliability design features.

Principle 4
Pilot your protocol/order set on a small scale before attempting wide implementation. Inevitably there will some glitches with your initial order set and/or algorithm. It's best to "fail faster" by piloting on a small scale, so you can get the glitches out of the way before you implement your protocol more broadly. The pilot can be as simple as a paper algorithm you ask 3-4 doctors to use or trying the order set on one ward.

Principle 5
Monitor the use of your protocol and order set: expect variation from the protocol and learn from it. Reduce variation from your protocol over time. Rolling out the protocol is really only a beginning. You need to learn from variations in your process. Why isn't the order set being used in some areas? Can we integrate it into other heavily used order sets? Which service needs our focused educational efforts? Which patients just don't "fit" with our protocol - can we change the protocol so that it fits more patients and situations? The idea is to squeeze variability out of the process while retaining variation based on tailoring to accommodate the patient.