The breadth and depth of knowledge in Medicine has meant that there are not only wide varieties of specialties and disciplines, but also many subspecialties. If you are a member of a physician group with highly skilled and talented doctors, then more than likely your group consists of many subspecialists. For example, Cardiology groups may consist of Echo cardiologists, Nuclear cardiologists, Imaging cardiologists, Transplant cardiologists, etc.  If you are a radiologist in a radiology group, you might be a Mammographer, Interventional radiologist, Neuro-radiologist, Diagnostic radiologist, etc.  You not only are an expert in your subspecialties but also have an interest and enjoy being in your area of focus.

However, if your responsibilities also include the administrative function of building and maintaining the schedule for your group of subspecialists, then you may experience little joy. In fact, you might spend a significant part of your week making schedules when you could be practicing your subspecialties!!!

By using a very simple scheduling example, this article answers the following questions:

  • Why does scheduling matter? Even in the simplest of examples, demand for doctors can be left unmet even when there are sufficient doctors available. Given the severe shortage of supspecialists, it makes sense to see how we can optimize the very valuable physician resources.
  • Why is scheduling challenging for both schedulers and most physician scheduling software tools? Scheduling is like playing a chess game that requires thinking many steps ahead. Often you may need to sacrifice a pawn to deliver a checkmate. Scheduling with such foresight is difficult not only for people but also for scheduling software products without sophisticated techniques like Artificial Intelligence.

An Example of Scheduling Physicians with Subspecialty Skills
Scheduling is complex. To simplify as much as possible, we will schedule a very small radiology group, Simple Radiology Associates, LLC (SRA).

table2

Scheduling this small group means replacing the question marks in the table with qualified doctors while meeting the following restrictions:

  • SRA has two subspecialty tasks, Neuro and Mammo, which need to be covered at just one site.
  • SRA has three distinguished (nobelaureate) doctors. Dr. Stoll subspecializes in Neuro, while Dr. Sharp subspecializes in Mammo. Dr. Krebs, a versatile radiologist, covers both Neuro and Mammo.
  • All the doctors need downtime. They will indicate to us when they must be on vacation or have a day off before we make the schedule.
  • We focus on scheduling just one day.

The qualified doctors are listed as either first choice or second choice. A doctor might gain their first or second choice status because the doctor or the scheduler explicitly expressed this as a preference or it could be dynamically/implicitly determined based on a conditional scheduling rule/requirement. As an example of the latter, a doctor who is ordinarily first choice may become a second choice during the scheduling process as a result of her having reached her work quota for the week/month being scheduled. Another example:  A doctor who is ordinarily a first choice becomes a second choice because she was on call the previous night.

Now let’s consider the various cases:

Case 1: Dr. Stoll on Vacation:

table2

This case is simple. Since the primary choices for each position are available we schedule Dr. Krebs for Neuro as a first step and, as a second step, assign Dr. Sharp to Mammo.

Case 2: Dr. Krebs on Vacation

table3

This case is also relatively simple. The only wrinkle is that we cannot schedule Dr. Krebs onto Neuro even though he is the primary choice because Dr. Krebs is on vacation. Dr. Stoll will take the Neuro slot while Dr. Sharp covers Mammo.  When a typical physician scheduling software vendor claims to auto-schedule subspecialty groups, their claim might be based on handling the above two cases only.

Case 3: Dr. Sharp on Vacation – A Myopic Choice

table4

Now we finally get to the most interesting case, the case of Dr. Sharp being out on vacation. When scheduling administrators schedule a task, they schedule the work slots, one-at-a-time, by deciding the physician for the current slot based on best immediately available options.

In this example, we might schedule the Neuro slot as a first step. Since the primary doctor, Dr. Krebs, is available, one would make the obvious choice of selecting Dr. Krebs for Neuro.  So far, so good. It’s not until we come to the Mammo slot that we realize our backs are against the wall because no one is available to cover that task. Of the two doctors qualified for Mammo, Dr. Sharp is on vacation and Dr. Krebs has already been scheduled for Neuro (cannot double-book her).

This same myopic, one-step-at-a-time process that got us into trouble is also what trips up many physician scheduling software tools; such computer-based tools just take us along the same wrong path faster! At least schedule administrators would recognize the need to backtrack and make the correction on the Neuro slot by choosing the sub-optimal second choice, Dr. Stoll, to allow us to finally schedule Dr. Krebs for Mammo.

With hindsight of 20/20, we have the final table that shows the optimal/correct schedule.

Case 3: Dr. Sharp on Vacation – Optimal Schedule

table5

To avoid having to backtrack and re-do portions of the schedule, we needed to think several steps ahead, much like a chess game. Also, like a chess game, we may need to sacrifice a pawn in order to produce the desired outcome of delivering a checkmate. This sort of forward thinking led us to settle for Dr. Stoll, a secondary choice, on the Neuro slot in order to not leave the Mammo slot uncovered.

This Case 3 example is one that any scheduler could easily solve because it did not require backtracking too many steps but in actual schedules, with 4 or more doctors and many more tasks, it gets tricky to do by hand because a poor choice at any scheduling step may not be recognized until hundreds or thousands of steps later. With group sizes of 20 or more, it’s almost a certainty that schedulers and typical software tools would miss opportunities to build an optimal schedule, resulting in one or more bad options below:

  1. Hire locums to cover uncovered tasks, at a cost in thousands of dollars per day per locum.
  2. Negotiate with physicians on vacations they have already planned. Good luck.
  3. Compromise patient care.

A fourth option that we at Lightning Bolt recommend is to create the best schedule possible using a smart software system such as our scheduling system built on Artificial Intelligence.