Excellent service processes are fast processes — this is the universal truth of service operations. Just as the water level in the bay hides the rocks from visibility, high levels of inventory mask underlying problems in manufacturing or logistics. Extending the analogy to service, long process times mask all kinds of service inefficiencies, such as incorrect order entry, insufficient resources, and errors in processing individual steps in the service journey. Increasing the velocity of a service process is akin to lowering the water level in the bay in order to expose the “rocks” that need to be tackled. Speed really does matter.
A medical appointment — really any appointment — can be divided into two parts. The value-added time is the time the patient is receiving “hands-on” treatment (e.g., vitals, talking with the provider, having a procedure). The non-value-added time is the time spent waiting for something to happen. In general, 70-80 percent of a patient appointment is spent waiting.
Increasing velocity requires an obsessive focus on making the non-value-added time go away; it is NOT about making the value-added steps go faster. Let’s dive deeper into the above example. If you were to speed up the value-added time (in red) by 2x, you’d reduce the total appointment time by five minutes. However, if you were to cut non-value-added time (in grey) in half, you’d shave 25 minutes off the total appointment time. In fact, there’s very little value-added time to be saved at all; for example, measuring vitals takes as long as it takes, and nobody would suggest that a provider spend less time with a patient. However, reducing time spent at arrival and check-in could cut 10 minutes; placing the patient immediately into an exam room to measure vitals could save another 10 minutes; and ensuring adequate staff are on-hand to eliminate checkout delays could reasonably save five minutes more.
Sounds easy, right? Just make people wait less and all will be good. Unfortunately, healthcare service delivery is particularly complex. One cause for this is that each person has a unique journey across the various treatment services (such as the clinic, labs, infusion, and imaging). Another is each treatment is highly variable in duration and depends on multiple factors. Yet another factor adding to the complexity of healthcare service delivery is that each treatment seems to have a “rush hour” when the incoming volume peaks, thereby causing delays to all downstream services. The resulting “domino effect” creates long wait times for patients and frustrates the staff who are trying to readjust the “Tetris blocks” in real time to get through all of the treatments that they are scheduled to complete that day. And like Groundhog Day, the whole cycle repeats itself the very next day!
The bulk of the process improvement work, regardless of the exact methodology used (such as lean or six sigma), accepts the sequence of scheduled patient appointments for the day as a given and then tries to improve the flow. Some ways to do this are making sure that orders are correctly entered and that the patient files are correctly assembled and sent to the right place at the right time. Unfortunately, the battle has been lost before it even began. The patient appointments scheduled for a given day were made as a sequence of individual negotiations between a scheduler and a patient without any regard to the portfolio of appointment durations that would be active at the time of the proposed appointment. Absent any formal attempt at mathematical optimization of the appointment sequence, it is in random order which will automatically be very far from the optimal sequence of appointments. Even the best poker player in the world can only do so much if they are consistently dealt a bad hand of cards — the same is true for the front line in healthcare service delivery that has to do the best it can with an appointment calendar that was assembled in a suboptimal manner.
Using mathematical optimization to create the appointment ensures that sufficient resources are at hand when the patient arrives, thereby matching the demand and supply patterns within short segments of time (e.g., 15-minute chunks). Applying data science and machine learning to schedule optimization ensures that the schedule is “resilient” — which means that even when the inevitable add-ons, delays and no-shows occur, the template will ensure that there is adequate capacity available to absorb the load. By synchronizing supply and demand based on the needs and constraints of each appointment, providers can increase velocity, leading to decreased patient wait times.
As first published in MedCity News.