Over a year ago, I wrote a blog about receiving a Covid-19 test and how simple process changes could be made to increase the throughput of the system—that is, to test more potentially infected people in less time without any additional resources, and without making anyone work any harder or faster.
A few weeks ago, I found myself in another Covid-19 line up. This time for a booster (third) shot. Since the Omicron variant has increased public anxieties about contracting the virus, people have been rushing to the hockey arenas (yes, I live in Canada) where public health authorities have set up temporary injection sites.
Unfortunately, just as when I had a swab shoved up my nose last year, I had ample time throughout the booster shot process to contemplate how to make the job of sticking needles into people’s arms more efficient. In fact, it took me a total 68 minutes from the time I entered the arena to the time I left it. To be clear: I’m NOT saying that the people working at the site are lazy or slow. On the contrary, they were all working hard. And they were polite, kind, and friendly too. But I feel the design of their process could be enhanced both for their benefit and for their customers’ (vaccine recipients).
The process for getting a shot, as I experienced it, has four steps to it:
- A symptom screening station, where they ask you if you’re feeling unwell, etc. (1 minute).
- A check-in process where they scan your health insurance card and verify personal data (1 minute).
- The injection itself, where they inform you of the risks of being vaccinated, jab you, and then explain common post-injection side-effects (10 minutes).
- A check-out process to ensure you received a confirmation of your vaccination status by email (5 minutes).
I’ve made a simplified diagram in Figure 1. By my count, there were 45 employees involved and I calculated they operated at a rate of about 150 shots per hour.
What struck me as the biggest opportunity in the vaccination process was the large amount of time devoted to waiting inside the process. I can somewhat understand the waiting at the front end, outside the process, where I waited in line for 15 minutes despite having an appointment, since there is more demand for shots than there is capacity to administer them right now, but any significant waiting in between steps is unnecessary and only serves to limit the productivity of the process.
Think of a popular rollercoaster ride at an amusement park. There is a fixed number of trains and maximum number of riders on each train. Each train takes a fairly predictable and repeatable time to load, complete its trip, and unload. If the arrival rate of ride-seekers exceeds the exit rate of ride-takers, then there will be a growing line up at the front of the ride. But no matter how long you might wait in line, once you get on board the ride you do not expect to stop moving until the ride is over. It’s designed as a continuous process. It would be completely absurd if the ride stopped mid-way through its run and required you to debark and wait on a platform beside the tracks for ten minutes before another empty train came by to pick you up to complete your ride. Yet for vaccine injections, and so many other business processes that are fragmented into discontinuous steps, such delays are not considered absurd at all. Waiting and delays are considered just a “normal” part of the work, even though they negatively impact productivity and can be reduced or eliminated altogether.
So how could we make the vaccine injection process more efficient without any additional staffing or technology, and without making anyone work harder or cutting corners on health and safety protocols? I would approach it in four steps:
- Eliminate unnecessary steps
- Shorten the duration of steps by eliminating unnecessary activities.
- Balance the throughput rate at each step.
- Repeat steps 1-3 forever
We can start with two of the easy pickings. First, the symptom screening station at the beginning is simply not necessary. I realize that some people might be feeling unwell, fear they might have Covid-19, and believe that a booster shot might alleviate their symptoms. This is, of course, false—the booster shot is a preventative treatment, not a therapeutic one—but I understand that a small minority of people might believe otherwise. Nonetheless, public health could handle this symptom screening in different, more efficient ways, such as getting customers to formally agree (via an electronic checkbox) to the rules (no entry if symptomatic) when they register for their vaccination appointment online. The rules could be reinforced again in multiple ways such as signage, handouts, temperature checks, text messages, audio announcements, and/or video displays while customers wait in line at the front-end of the process. Does it really require a live human being to ask every single visitor whether they have a fever or have traveled outside the country in the last 14 days? After nearly two years of living with this pandemic, we’ve all memorized these rules for entering any non-residential building, have we not? If someone is trying to sneak in a shot while they are symptomatic, they likely know full-well they are breaking the rules and are deliberately lying to the screeners. This screening step seems like a highly ineffective use of resources to me. If this step really, truly cannot be eliminated because of some legal liability issue, then it should, at the very least, be combined with the second step, the check-in, to avoid the queue between the symptom screening and the check-in.
Secondly, the post-vaccination waiting area is designated as a place where customers are requested to sit for 15 minutes to see if they have an immediate adverse reaction (e.g. an allergic reaction) to the vaccine. If they do not have any adverse reactions in 15 minutes, they can check-out and go home. Serious adverse reactions are very rare, especially so because customers are receiving their third dose—if they did not have any severe reaction after the first two shots, what are the chances on the third? This seems like an overly cautious step. Indeed, the woman who seated me in the first waiting area (prior to vaccination) told me sotto voce that it could be considered optional and that some customers who were in a rush had skipped this post-shot wait. As usual, the people working in the process have the best ideas. So why not make it optional, at least for those who have had no previous allergic reactions to the same type of vaccine and who do not have any medical conditions that might elevate their risk of a reaction. The goal of vaccinating as many customers as fast as possible would seem to outweigh the very small risk of an allergic reaction.
Why did it take me 10 minutes to get a shot that lasted no more than 5 seconds? First, the vaccinator had to go through a long list of questions asking me if I have health conditions (e.g., heart problems, blood problems, etc.), allergies to specific ingredients found in vaccines, and a bunch of other stuff—i.e., all the “informed consent” protocols. I appreciate that this is important and mandatory, but could this info have collected online ahead of time when I made my appointment? Perhaps I could have been sent a digital form on my phone 15 minutes before my appointment? Maybe there is some information security risk or other legal obligation for the vaccinator to gain my verbal consent at the time of injection that I’m not aware of. If not, it’s worth considering offloading this part of the process onto the vaccine recipient prior to their visit.
The bigger opportunity at this stage of the process was that I had to pull out my phone and retrieve my vaccination history and show it to the vaccinator. This strikes me as information that she should have had at her fingertips, given that I had already handed over all my personal info at check-in. She even validated my personal info again with me on her iPad—so why not have my vaccination records scanned at check-in and shared with the vaccinators?
In any case, once she saw my previous vaccinations, she decided to inject me with the Moderna vaccine. This posed a problem because she only had Pfizer vaccines at her table. She had to get up and go to a colleague’s table to get some Moderna ones. This took a few minutes.
If the vaccinator had had my vaccination history on her iPad as soon as I told her my name, and she had had both brands of vaccine handy, I think this step could have easily taken only 5 minutes, including the time it took her to ask me the list of questions and to re-verify my personal info.
There was another obvious opportunity to shorten the process later, at the check-out stage. After waiting 15 minutes, I had still not received my emailed proof of vaccination, even though both the check-in and vaccination steps had confirmed that the email they had on file was correct. Someone, I think, forgot to send it to me, or perhaps it was a system glitch. Whatever the case, I had to wait 5 minutes for the check-out employees to fumble through a database and send me my email before I could leave. This check-out step should take a maximum of 30 seconds if the email is received and should only take longer for those few customers who require a printed paper copy. A little bit of error-proofing in this process could go a long way to shortening this step for people like me.
The biggest opportunity of all in this entire process is the large waiting area between the check-in and the injection steps. This area is comprised of a 5 x 5 array of chairs, arranged in clusters of 2 or 3, with all the clusters spaced 6+ feet apart for the sake of social distancing. It occupies at least a quarter of a full-sized hockey rink and seats roughly 50 people, although there were probably only 30-35 customers waiting there when I visited, since some people (like me) had come alone to the clinic. I—and everyone else with me—spent approximately 20 minutes waiting here. There was a public health worker (PHW) whose job it was to seat customers in designated chairs, like a hostess at a restaurant leading you to your table and inviting you to sit down. She also had the responsibility of releasing us to the vaccination area on a first-in-first-out (FIFO) basis. She had a methodical system worked out where she rotated through the array of chairs sequentially to ensure she maintained the FIFO order.
Standing in front of the vaccination area was yet another PHW scanning for open injection stations. The vaccinators would raise a big red circle made from two red plastic plates glued to a stick, like giant lollipops, to indicate to the PHW in front of the vaccination area that they were free to inject another customer. This PHW would then signal with her fingers to her colleague in the waiting area how many could now be released to the vaccination area, simulating a low-tech version of a check-out line at Walmart.
While waiting rooms are a build-in fixture of many medical services, and so have become normalized, this area is completely unnecessary in this case and should be eliminated. The only reason the waiting happens is because the check-in step is checking-in customers faster than the injection step is vaccinating them. A basic law of processes is that queues (aka bottlenecks) will always form when a prior step operates faster than the step immediately following it. In this case, the vaccination step is the slower step. How do I know? Well, as I mentioned above, it took a full 10 minutes for the vaccinator to stick a needle full of 50 micrograms of Moderna Covid-19 mRNA-1273 in my arm for a few seconds. The check-in step, in contrast, took about 1 minute. This does not mean the vaccination step is currently operating exactly 10 times slower than the check-in step because we haven’t yet accounted for the number of people working at each step. But given that there is a 20-minute wait between steps, we can assume that the vaccination step is operating at a rate that is slower than the check-in step. To know exactly how much slower, we can calculate the throughput rate of each process step.
Calculating the throughput rate for any given process step involves some simple math. I know many people recoil at the word “math”, but it really is quite simple. I took the time it takes for one person (me) to go through this step and then divided it by the number of employees performing this step concurrently. The throughput rate for the vaccination step ended up being one vaccination every 24 seconds. Be careful not confuse the throughput rate and the lead time. The throughput rate does not mean that it only takes 24 seconds for one customer to go through the step—remember, it took me a full 10 minutes (the lead time). What it means is that the process will output one vaccinated individual, on average, every 24 seconds, under current conditions. It’s a measure of a process’ efficiency, not a measure of an individual’s experience. Remember, there are multiple customers being “processed” (vaccinated) by multiple vaccinators simultaneously. By calculating a step’s throughput rate, we can compare the relative efficiency of different process steps that are part of the same overall, end-to-end process, but which have different durations and different staffing levels.
Following from this, if we could pace the arrival rate of customers at the vaccination step to equal one every 24 seconds (around 13 customers every 5 minutes), to match its capacity, there would be no real need for any waiting area at all. Unfortunately, the check-in process is feeding the vaccination step much faster, at a rate of 1 person every 5 seconds. This lack of balance (5 seconds compared to 24) in the throughput times of the two separate processes is causing the additional waiting time. Even if we could cut down the average time it takes to carry out the injection step to 5 minutes, as I suggest above, the process is still very unbalanced (5 seconds compared to 12).
Well-designed processes are balanced processes. This means that all the process steps are processing the work (or the customers, in this case) at the same rate. To eliminate the bottleneck at the waiting area, we need to either slow down the throughput rate of the check-in process or speed up the throughput rate at the vaccination step, or both. For instance, if we could move the 3 staff from the symptom screening area (which we’ve now eliminated) to the vaccination area, this would immediately increase the latter process’ throughput rate to one injection every 11 seconds (assume the improved 5-minute processing time). We could then potentially move some people working in the check-in area to work in the vaccination area. It turns out the optimal number to move is 5, thereby leaving the check-in step with only 7 people. This would effectively slow the throughput rate of the check-in step to one customer every 8.6 seconds while accelerating the rate of the vaccination step to one customer every 9.1 seconds (remember, less time means a faster process). Now they are effectively balanced since both are operating at one customer approximately every 9 seconds. The balancing of the throughput rate of two process steps completely eliminates the need to have any waiting area in between them. This saves 20 minutes of the customers’ time and allows more customers to be vaccinated in less time with the same number of resources working at the same pace.
As a bonus, now that we’re able to eliminate the waiting area, the 2 people working to seat and release customers in the waiting area could possibly be added to the vaccination team as well, improving (and balancing even further) the throughput rate at that step from a customer every 9.1 seconds down to one every 8.6 seconds.
This transfer of staff from other roles to vaccination assumes, of course, that the PHWs in other roles are certified to administer shots. The current regulations where I live (province of Ontario) require vaccinators to be a certain type of healthcare professional (e.g., nurse, physician, pharmacist) and to have completed an approved course in injection safety. These requirements may limit the amount of flexibility the process has. Do the people working at the symptom screening or check-in steps have these qualifications? If not, can they be obtained quickly? It would make sense, in the face of the current pandemic, for provincial health authorities to loosen the requirements somewhat to increase the number of people who can safely inject others with a Covid vaccine. Some provinces have allowed veterinarians, physiotherapists, and paramedics to administer the shots too. Some provinces have abbreviated the course requirements to increase the number of people able to administer the shots. Given the current shortage of healthcare professionals—they are quitting in droves due to the burnout and stress dealing with Covid for two years— these trade-offs seem warranted.
And consider the benefits of being able to move qualified staff around a process as I’ve suggested. In the best-case scenario, without the symptom screening step and dispensing with the need to hang around for 15 minutes after the shot, every vaccine recipient could fly through this process in 6.5 minutes: 1 minute to check-in, 5 minutes to get vaccinated, and 30 seconds to check the receipt of the email confirmation. Add in a couple minutes to walk from one area to the next and we’re looking at 9 minutes in total. Even if there were still a 15-minute wait at the front-end, we’re looking at a 24-minute stay instead of a 68 minute one—more than a 50% reduction in time spent getting vaccinated. And from a productivity point of view, a throughput rate of 8.6 seconds means a near threefold increase the number of shots administered in the same time, with no net increase in resources. See Figure 2.
But why stop there? There’s still lots of opportunity to improve this process.
For instance, why do we have to wait 15 minutes at the front-end of the process? Unlike an amusement park ride, the booster vaccinations times are scheduled exclusively by appointment. By maintaining a balanced process with a throughput time of 8.6 seconds, we can predict with fairly good accuracy how many vaccines we can deliver into arms in a given time period. It turns out it is about 35 customers every 5 minutes. This means that public health authorities can schedule 35 customers to enter the arena every 5 minutes, or one person every 8.6 seconds, on average. Since the scheduling now balances the arrival rate with the throughput rate of the process, there should be, on average, only a 2.5-minute wait at the front of the process. So now everyone can move through the process even faster: 2.5 minutes waiting, 1 minute to check-in, 5 minutes to get vaccinated, 30 seconds to check-out: 9 minutes in total. Allow another 3 minutes to move around between process steps, and we’re looking at an end-to-end duration of only 12 minutes—an 82% reduction in elapsed time for each vaccination recipient. At 35 customers every 5 minutes, this means 420 shots/hour. See Figure 3.
Now I know that you’re probably thinking I’m being unrealistically utopian and that I know nothing about healthcare. You’re thinking I’m being reckless with safety protocols, and I should go back to being an improvement nerd about processes that don’t involve vaccines, sterilization, or any understanding of human physiology. Ok, you have some good points. So, let’s say we cannot move anyone else into the vaccination step. Let’s say we cannot dispense with the 15-minute waiting period post-vaccination. We can still reduce the waiting by balancing the process steps. We could reduce the number of people employed at the check-in to 5 and maintain the 25 in the vaccination area to achieve a balanced throughput of one person every 12 seconds (about 33% slower than the “utopian” version). The scheduled appointments would allow 25 customers into the arena every 5 minutes, allowing for 300 shots/hour, a 100% increase over the current situation, and with a 27% reduction in staff. The lead time would be about 27 minutes—a 60% improvement. See Figure 4.
I hope you’re now asking yourself what would happen to the 12 “excess” PHWs (3 from the symptom screening; 7 from the check-in; and the 2 seating coordinators). Process improvements should never lead to layoffs or reduced hours. That is why I am a big fan of cross-training to allow for a more flexible workforce, improved process efficiency, and career growth for staff. But, in this case, I think these 12 people could be redeployed to other understaffed areas of Ottawa’s healthcare system. No one should be out of work, and it would be a more effective use of resources.
Eliminate, shorten, and balance. Despite the dramatic improvements that we can achieve by approaching our processes with these three guidelines in mind, there will always be constraints, new conditions, and new challenges that arise. And we can always do better. Think of how we might reduce the check-in process to less than 60 seconds or the vaccination process to less than 5 minutes. Think of how we might reduce the time and effort it takes for customers (especially pertinent for the elderly) to walk between stations. There is no end to the improvements we can make. If we embrace continuous improvement, we can achieve better public health outcomes for all.
 This is calculated based on the throughput rate at the vaccination stage being 1 customer/24 seconds (more on this below). 3600 seconds (1 hour)/(1 customer/24 seconds) = 150 customers/hour.
 Simply calculated by the number of riders on a train divided by the time interval between trains arriving at the unloading area.
 Lean practitioners will recognize this as a kind of crude kanban method to “pull” customers into the next process step.
 What lean people usually call “cycle time”
 There were 25 injection stations, and it took me 10 minutes. So, the process’ throughput rate is 600 seconds/25 vaccinators or one vaccination every 24 seconds.
 12 people performing a process that takes 60 seconds = 60/12 = 5 seconds.
 For the check-in process, the throughput rate is now calculated as 60 secs/7 staff = 8.6 seconds; for the vaccination step, it is 300 secs/33 staff = 9.1 seconds.
 The hourly vaccination count can be calculated as 3600 seconds / (8.6 seconds/customer) = 419 vaccinations/hour
 If you want to do the math, divide 5 minutes (300 seconds) by 8.6 seconds/person.