In this case study, Healthcare Systems Design expert, Daniel Custódio, provides a solution for healthcare executives to create hospital capacity and reduce COVID-19 related deaths.
Click here to view a Youtube video on this case study.
David C. Klonoff, MD, FACP, FRCPE, medical director of the Mills-Peninsula Medical Center (Sutter Health) Diabetes Research Institute in San Mateo, California, recently led a study showing that patients afflicted by COVID-19 with diabetes and uncontrolled hyperglycemia have a higher mortality rate (28%) than those without these two conditions (6%).
Although the research is preliminary, there is strong evidence to suggest that better glycemic management could reduce the length of stay (LOS), freeing up beds for those in need, and leading to fewer COVID-19 related deaths. This begs the question for any healthcare executive, how can healthcare systems better manage diabetic patients with uncontrolled hyperglycemia?
In 2019, the Kōhei Group was hired by a healthcare system to help them improve outcomes. Glycemic management was one of the areas we identified for improvement. On any given day, up to 25% of patients in a medicine unit could be diabetic. Glycemic mismanagement was contributing to patient harm in the form of hyper and hypoglycemic events and, as a result, longer-than-expected LOS. Potential causes of these events were as follows:
- The administration of too much insulin or diabetes medication
- Not eating enough
- Postponing or skipping a meal or snack
- Increasing physical activity without eating more or adjusting your medications
- Missing administration of insulin or glucose-lowering medication
- Eating the wrong foods
- Eating too much food
- Decreased activity
Through direct observation and data analysis, we confirmed several contributing factors.
- Insulin was administered well after a patient’s blood glucose level was recorded (only 23% of patients were receiving insulin within the recommended timeframe of their blood glucose level being recorded). As a result, the dosage being administered most likely was not correct.
- Two nurses were needed to administer insulin: one to administer and the other to confirm the dosage was correct. However, the administering nurse would usually have a hard time finding another nurse and, thus, would delay administration.
- Meal delivery was not synchronized with insulin delivery, thus, delaying when patients ate.
- Patients were not sufficiently monitored to ensure they were eating the appropriate amount of food and getting the right amount of physical activity
A Model for Treatment
Once we understood the problem, we turned our attention towards the solution. The following model is one used by the Kōhei Group to create fail-proof systems for any aspect of care delivery:
In this model, we begin at the center circle. In our example, this would translate into a way of delivering care that would guarantee the reduction of hypo and hyperglycemic events. Most solutions stop here, but as we know, defining a better way of working does not mean that it will happen.
The second component of our model, Sensory Management, consists of three key sub-components:
- Clear accountability
- Alerts to increase the likelihood frontline staff deliver care per the standard work
- Feedback on performance
Frontline staff begins their shift by referring to a visual of their assignment. This creates clarity regarding what they are accountable for. As they work, alerts provide them feedback when they are at risk for not delivering care to the standard work. This can be in the form of an audio, visual, or tactile alarm. The third subcomponent is also where the third component of our model, Self-Accountability, is introduced.
At appropriate intervals throughout their shift, the frontline staff takes efficient pauses to reflect on how they are performing to the standard work and adjust. Equally as important, they capture obstacles (in the form of a Pareto Analysis) that got in their way of delivering care per the standard work. A key principle to make this system work is known as a focus on the process. In other words, we assume obstacles are not created by human error, but rather, are a direct result of a poorly designed system. The statistician, W. Edwards Deming, confirmed through his research that this is, in fact, the case. Reinforcing this principle creates a no blame work environment, which, in turn, increases the likelihood of frontline staff sharing obstacles that get in their way .
The next component of our model, Observation, is the glue that holds everything together. In this model, management has a specific role as a coach. At appropriate intervals, they should go to observe the work directly. When they do, they are going with eyes for the following:
- Visual management is up-to-date, specifically related to clear accountability and feedback on performance
- Frontline staff are taking efficient pauses to reflect on how they are performing to the standard work and adjusting
- When standard work cannot be followed, obstacles are being captured
At first, this will seem abnormal and will not be welcomed. That is why it is critical it is done with the mindset of no blame. When observation is followed up with our last component, Problem Solving, and managers remove obstacles, the frontline staff will soon welcome the observation.
Depicted below is our system’s solution to diabetic care at a high level.  Steps 3, 5 and 6 represent the standard work for care delivery, with steps 1, 2, 4, and 7 providing systems support to ensure the probability of delivering care in the prescribed manner is increased.
The first step (step 1) requires the charge nurse to update a planning board for the next shift. The information updated includes indicating which patients are diabetic and the order in which care delivery will be executed. In addition, the charge nurse will walk the unit floor and place magnetic strips and timers, as seen in the picture below, on door frames of diabetic patients.
Next (step 2), a NST, commonly referred to as a Tech, will go to the planning board at a set time and see their assignment for recording patient blood glucose levels. With clarity on their assignment, they will begin to execute (step 3).
Prior to entering the patient room, the NST starts the alarm, which will go off at 20 minutes (a warning) and at 30 minutes (a hard stop) to keep everyone on track to deliver insulin within a safe range of time. They then record the patient’s blood glucose level per their standard work. Should any obstacles get in their way, they collect obstacles at a point-of-use Pareto chart. When they are complete, they move the alarm from the red zone (“finger-stick”) to the yellow zone (“insulin”) to indicate their job is done and the next job can begin.
Next (step 4), the NSTs, RNs and an assigned staff member from Food Service Deliver begin their shift. They meet members of the previous shift at the planning board to provide clarity on their team assignment. Another best practice that was not adhered to prior to our transformation work was bed-side hand-off with members from the previous shift updating members of the next one. By creating a team that worked in concert at bedside we could implement that best practice and overcome the obstacle of a nurse not being able to find a second nurse for the administration of insulin (step 5).
Upon completing bed-side hand-off and the administration of insulin, the team moves the alarm from the yellow zone to the green zone to indicate food delivery can proceed (step 6).
Throughout this entire process, the Charge Nurse is responsible for ensuring the frontline staff are attempting to follow the standard work and, more importantly, if they cannot, capturing reasons as to why. The alarms should trigger the Charge Nurse to go and observe but also to ensure no insulin is administered after 30 minutes. Without the Charge Nurse, this system fails.
It is important to note that for steps 3, 5, and 6, best practices were researched to provide guard-rails for frontline employees to develop standard work on how to best deliver care. Each time an obstacle got in our way, we conducted controlled experiments to overcome those obstacles. When those experiments yielded positive results, we refined the standard work in the spirit of continuous improvement. It is also important to note, that the same systems thinking expanded to the practice of patient hour-by-hour rounding to ensure other aspects of care were addressed to positively impact the diabetic and non-diabetic patient journey.
When we began our work, compliance to the best practice of insulin delivery was only 23%. As you see in the chart below, when we launched our new system, we were able to improve immediately by 22%.
Some may attribute this to the “Hawthorne” effect, however, as the graph shows, week-over-week, we refined the system until we were passed our target of 85%. As a result, the outcome metrics the hospital was trying to achieve also saw significant improvement. In the two units where we rolled this system out, one reduced its LOS by .49 days and the other by 1.6 days, which had a tremendous impact on the bottom line. Harm was also significantly reduced when compared to other units and was at zero for many months. It is also important to note, that there were days when the goal for insulin delivery was met 100% of the time and that the weekends brought down our performance due to lack of training.
If this could be achieved in 8 weeks, what could be achieved in 52 with continuous improvement as a focus? To take it one step further, if a healthcare system were to apply this systems thinking to all their core care delivery, what impact would that have? If every healthcare system in the US did so, what impact would that have? It is clear from Dr. Klonoff’s research that thousands upon thousands of lives could have been saved by now if healthcare systems implementing this kind of thinking prior to the pandemic. The good news is that a disaster always presents the opportunity to improve. That time is now.
UPDATE: Since this was published, we have conducted an analysis that shows over 23,000 patients who died with COVID-19 could have been saved if this solution were in place at each US hospital.