If you’re of a certain age, you remember Tetris, a tile-matching puzzle video game originally designed and programmed by Russian game designer Alexey Pajitnov. Back in 1984 – before “viral” anything was a thing – millions of people became obsessed (addicted?) to this simple game, which challenged players to stack blocks of different shapes and sizes as efficiently as possible. Stacked improperly, the blocks would reach to the top of the screen – game over. Stacked optimally, the game could go on for hours.
What does this have to do with infusion scheduling? It turns out scheduling patients with varying treatment lengths is a lot like Tetris, only the odds of winning are stacked (pun intended) much more against you for a variety of reasons, including:
- Most centers schedule patients on a first-come, first-served basis – much like a corner hair salon – with no regard for who’s in the chair next to them, before them, or after them.
- Under the noble goal of patient satisfaction, schedulers let patients dictate their appointment time. Since most patients want to receive treatment between 10am and 2pm and chairs are a finite resource, exceeding capacity in the middle of the day is common. Hoping for no-shows is not a successful scheduling strategy!
- With the randomized arrangement of appointments, and the risks of patients arriving late, having an adverse reaction, nurses calling in sick, the pharmacy running behind, the clinics running late, it doesn’t take much to completely blow up the day’s schedule…..only to have it happen all over again the next day
- Simple spreadsheets or traditional EHR approaches are not designed to create an optimal solution for scheduling appointments.
Frankly, the math involved in creating an optimized daily schedule is daunting. Let’s look at an example:
- Take a 35-chair infusion center that operates eight hours per day treating five types of appointments – 1 hour, 2 hours, 3-5 hours, 6-8 hours, or 9 or more hours.
- Assume four sets of patients can start their treatment at 10-minute intervals. That’s 256 possible start times or “slots” per day.
- The number of possible ways these patient appointments can be arranged is a number with over 100 zeros behind it. By comparison, if you were to use 1-gallon milk jugs to hold all of the oceans’ water, the number of jugs needed would have 40 zeros behind it. Compared to finding the best way to arrange these appointments, finding a needle in a haystack is a piece of cake!
So if the numbers are stacked against infusion center schedulers, how can they possibly arrange the roster of appointments – not just for today, but for every day forward for the next several weeks – in a way that allows the center to keep up with increasing patient volumes, to prevent excessive wait times, and to keep operational costs down? Unfortunately, in a problem with multiple key resource constraints, focusing on just one gives a sub-optimal answer at best, and has the potential to introduce major bottlenecks at worst. Scheduling to a nurse hides the chair utilization, so it’s possible to build a schedule that lines up patient starts well for the nursing team, but creates a major peak in chair utilization. Similarly, scheduling to a chair makes it difficult to plan a schedule that will work well for nurses.
On top of that, filling in schedules for a particular resource as each patient need arises means that the schedule is built through a series of isolated negotiations between a patient and a scheduler – without considering the rest of the patient demand that still needs to be booked. To end up with an answer that works efficiently across multiple constrained resources, you need to run a constraint-based optimization that comes up with a set of appointment options that account for the full expected demand, while ensuring that utilization of both chairs and nurses is optimal, and also spacing out patient starts so that the flow through check-in and pharmacy is smooth as well.
And what about pods? Splitting the problem into pods makes it easier to conceptualize the demand on multiple resources, but creating smaller groupings of nurse and chair resources limits efficiency overall, and locks you into patterns that may not work once the variability of the day hits – with some patients arriving late from clinics, others who have a bad reaction to a drug and need to stay in a chair longer than planned, and others who need to be urgently added to the schedule.
The key ingredient, of course, is data, more specifically EHR data. Inspired by the likes of Toyota and just-in-time lean manufacturing practices, data science and mathematics are changing the face of healthcare scheduling and, in effect, making healthcare more accessible to more patients. LeanTaaS data scientists are mining all scheduling patterns and possibilities specific to a center, as well as considering operational constraints across patient demand, practitioner and staffing schedules, and capital asset availability. From there, an algorithm optimizes a schedule for the center to serve patients more uniformly throughout the day, versus the peaks and valleys of traditional scheduling systems.
This mathematical approach to infusion center scheduling is already delivering impressive results; providers like Stanford Health Care, UCHealth, NewYork-Presbyterian, UCSF, the Huntsman Cancer Institute, Memorial Sloan Kettering and many others are accommodating a 15% average increase in volume, seeing wait times decreased by as much as 55 percent during peak hours, and reducing overtime hours by as much as 74%. Put that in real-world terms: A one-hour wait becomes 27 minutes. Who wouldn’t want a half-hour of waiting room time back? You might even be able to squeeze in one more game of Tetris.
Former Sr. Partner at McKinsey and Co.LeanTaaS
Latest posts by Mohan Giridharadas (see all)
- LeanTaaS to Showcase Impact of Predictive Analytics on Healthcare Operations at HIMSS18 - February 21, 2018
- UCHealth Improves Operating Room Efficiency, Lowering Wait Times and Improving Patient Experience, with LeanTaaS - October 24, 2017
- Health systems need to move beyond resource-based scheduling - July 27, 2017