How heavily should a firm use its resources? Resources — be they people, equipment, or facilities — are expensive so there is an obvious case to be made for keeping utilization rates as high as possible. But there is also something to be said for not pushing utilization too high. Many systems need some slack to work well. It is slack that allows firms to absorb the unpredictable or to address problems that go beyond immediate firefighting. That is the point of a recent Strategy & Business article (Cut Your Company’s Fat but Keep Some Slack, Spring 2014). The authors main point is that “slack is routinely undervalued.”
Here is the example given to lead off the article.
In 2002, the operating rooms at St. John’s Regional Health Center, an acute-care hospital in Missouri, were at 100 percent capacity. When emergency cases—which made up about 20 percent of the full load—arose, the hospital was forced to bump long-scheduled surgeries. As a result, according to one study, doctors often waited several hours to perform two-hour procedures and sometimes operated at 2 a.m., and staff members regularly worked unplanned overtime. The hospital was constantly behind.
Administrators brought in an outside advisor, who came up with a rather surprising solution: Leave one room unused. To many, this seemed crazy. The facility was already being squeezed, and now comes a recommendation to take away even more capacity? Yet there was a profound logic to this recommendation, a logic that is instructive for the management of scarcity.
On the surface, St. John’s lacked operating rooms. But what it actually lacked was the ability to accommodate emergencies. Because planned procedures were taking up all the rooms, unplanned surgeries required a continual rearranging of the schedule—which had serious repercussions for costs and even quality of care. The key to finding a solution was the fact that the term unplanned surgery is a bit misleading. The hospital can’t predict each individual procedure, but it knows that there will always be emergencies. Once a room was set aside specifically for unscheduled cases, all the other operating rooms could be packed well and proceed unencumbered by surprises. The empty room thus added much-needed slack to the system. Soon after implementing this plan, the hospital was able to accommodate 5.1 percent more surgical cases overall, the number of surgeries performed after 3 p.m. fell by 45 percent, and revenue increased. And in the two years that followed, the hospital experienced a 7 and 11 percent annual increase in surgical volume.
At a high level, I find the basic argument of this article appealing: An organization can be too lean and without extra resources it lacks robustness. Without slack, every hiccup becomes a catastrophe and various performance metrics suffer.
I also really like the St. John’s operating room story because it presents an interesting conundrum in queuing. First, the need for slack is one of the key points in queuing theory. The insight that at some point a small increase utilization can lead to a dramatic increase in waiting time is one of the first points we make in teaching MBA students about managing service systems. Good performance then requires some amount of extra capacity.
The challenge then is to identify smart ways of managing that extra capacity. One smart way is to pool capacity. There are economies of scale in managing queues. Systems with higher arrival rates are going to need more resources to hit a given service level (e.g, having the average wait below some target) but they will be able to do so at a higher level of utilization. Thus big systems are going to need more capacity but have a lower cost per transaction.
Now think about what St. John’s did. They gave up on pooling. Holding back one OR means that they are not pooling resources and thus one should expect worse performance.
There are, arguably, two reasons why sacrificing pooling actually made things better at St. John’s. First, there is the question of what is the right definition of performance. For most queuing systems, the emphasis is on avoiding delay — less time on hold at a call center is seen as improvement. Little or no delay is obviously an important goal for emergencies coming into a hospital but that is not necessarily the right way to think about scheduled surgeries. If someone is coming in for a procedure at 10:00AM on Thursday morning, the appropriate question is how closely did the surgery keep to this schedule. Waits before 10:00AM are not relevant because the patient was told nothing would happen before then. Delays after 10:00 are a different story since the hospital is now deviating from the promised start time.
A performance measure focused on schedule adherence then interacts with another consideration: Priorities. Few would argue that a patient in need of a life-saving procedure should not be given priority over patients for whom waiting is inconvenient but not life threatening. But that wreaks havoc with trying to have scheduled procedures start on time. If an operating room becomes available, there is every reason to give it to an emergent patient and bounce a scheduled surgery. Giving up on pooling then protects scheduled patients while improving service for emergent ones.
But wait there is more. Another smart way of managing the slack needed in queuing systems is reduce variability — say by having work arrive according to a schedule instead of having customers just randomly show up. However, I suspect that how St John’s was scheduling work was part of the problem. Check out this graphic from a study of a different hospital (Boston hospital sees big impact from smoothing elective schedule, OR Manager, Dec 2004).
These plots show the number of surgery patients being sent to either a post-surgery stepdown unit or a surgical intensive care unit.Unscheduled patients are on the left and a scheduled patients on the right. Note that for the unscheduled patients one day of the week looks pretty much like any other but there is a clear weekly pattern for scheduled cases — both in volume and mix of doctors doing the surgeries. This is a complete artifact of how the hospital allocates operating rooms to physicians and it can actually increase rather than decrease variability.
This is a possible explanation for how St. John’s was able to do more surgeries after dedicating an OR to emergency cases: Smarter scheduling would allow them to flatten out the schedule across days, assuring that capacity is not wasted.