The notion of improving operational efficiency is conspicuously absent from the healthcare debate — neither Obamacare nor any of the proposed GOP plans discuss the impact that a step-function improvement in efficiency could have on access to healthcare (through more capacity), quality of healthcare services (through reduced wait times for patients) or cost (through better utilization of scarce, expensive assets).
The opportunity of improving operational efficiency in health systems is plainly visible — a quick walk around any hospital or clinic will show the obvious symptoms. Waiting rooms fill with people while the very piece of equipment for which they are waiting remains idle. Patients spend 90 minutes or more to obtain a quick consultation or a check of their vital signs, which collectively require less than 10 minutes for a provider and/or their assistant to complete.
The traditional approaches of process improvement — using lean, Six Sigma or some other methodology — have run their course. In most cases, they only deliver modest improvement considering the enormous resource burden they place on the organization and the time it takes to show tangible results.
We are at a pivotal point in time. The convergence of multiple technologies makes it possible for us to drive a step-change improvement in operational efficiency that far exceeds anything that we could have imagined, even a few short years ago. Specifically, this includes:
- The massive digitization of patient data;
- Internet of things providing real-time data on the movements of patients, providers or assets;
- Emergence of machine learning and artificial intelligence;
- Democratization of predictive analytics from the ivory towers of academic institutions to organizations everywhere;
- Growth of massively scalable, secure cloud infrastructure;
- Ubiquity of smartphones and mobile apps.
And the list goes on. We have seen this phenomenon in other contexts, such as Uber, Snapchat, Instagram and many other “unicorns” could not have existed had the smartphone not become prevalent.
A Deep-dive Into Specialty Pharma
A specialty drug is a class of prescription medications used to treat complex, chronic or rare medical conditions. Although this classification was originally intended to define the treatment of rare, also termed “orphan” diseases, affecting fewer than 200,000 people in the US, more recently, specialty drugs have emerged as the cornerstone of treatment for chronic and complex diseases such as cancer, autoimmune conditions, diabetes, hepatitis C, and HIV/AIDS.
Combining lean thinking with predictive analytics, machine learning, advanced optimization algorithms and embedding it into scalable software products can drive dramatic improvements in the operational efficiencies of individual units in a hospital. Health systems are a complex interconnected network of individual units — labs, imagining departments, pharmacies, infusion centers, and operating rooms — each of which provides a specific service. In any interconnected network of units, it is much more effective to optimize the individual units before seeking to optimize the interconnections. Hence, UPS and FedEx will optimize their sorting hubs and warehouses as much as possible before worrying about their drivers driving faster.
This approach has yielded tremendous results. For example:
At 60 infusion centers across the country, patient wait times during peak hours have been reduced by 30 percent to 55 percent and the effective capacity of the centers increased by 15 percent 20 percent as a result of using models that can accurately predict the volume and mix of infusion treatments tailored to each infusion center for each day of the week.
Heijunka
These centers have incorporated center-specific parameters (such as capacity and staff) into an optimization algorithm that created tailored appointment templates for each hour of each day of the week that consistently deliver against the core lean principle of “heijunka” — or level loading — which reduces the wait time for patients while balancing the workload for nurses.
Approximately 100 operating rooms and about 200 surgeons have been able to improve block and room utilization by 5-7 percentage points by predicting the need for block time for individual surgeons and service lines. After they accurately estimated the demand for block time, algorithms then identify the right supply of blocks by uncovering patterns of underutilized, abandoned or late-release blocks to give each surgeon and service line the correct number of blocks of the correct length on the right day of the week. Other artificial intelligence algorithms automatically generate recommendations that encourage surgeons to release blocks that are not likely to be well utilized and even facilitate the “swap” with another surgeon using an OpenTable-like mobile application for block swaps. Blocks swapped in this way performed 12 percent to 15 percent better than blocks swapped using the conventional methods of schedulers sending out a flurry of phone calls, emails, faxes and voicemails to assign a newly available block to a surgeon or service line.
Predictive models
Predictive models from iQueue Labs have enabled the Emory Winship Cancer Institute to reduce the wait time in the lab at peak times from approximately one hour to less than 15 minutes. Emory has found that accurately predicting the volume and mix of patients (blood draw versus central line patients) at 15-minute increments for each day of the week makes it possible to correctly staff the number of phlebotomists and LPNs in order to virtually eliminate the wait time for patients. The result is not only a reduction in wait times in the lab; the improvement has also positively affected “downstream” services such as infusion treatments.
These are just three examples.
Hospitals have many other opportunities for solving operational challenges that plague them on a daily basis. Other examples include both “supply side” problems (e.g., critical assets such as CT/MRI scanners, blood testing equipment, personnel) where asset utilization and/or availability are challenges as well as “demand side” problems (e.g., labs, clinics) where accommodating walk-in or scheduled appointments in a timely manner is a challenge.
A well-functioning air traffic control capability, along with an effective airport operations function, has been able to unlock an enormous capacity for flights out of major airports. For example, Atlanta airport in the mid-1980s only had a few hundred flights per day. Today, they have several thousand flights per day. This was accomplished without a change in the airspace around Atlanta and only a modest increase in the number of runways (from four to five) over this time period. Like health systems they, too, have a very stringent requirement on safety — 99.9999 percent safety is simply not good enough since it would imply that society would accept a crash every few days in the United States.
A focus on using sophisticated “lean plus data science plus machine learning plus optimization plus scalable software” to unlock capacity and improve the throughput of the individual units within the hospital will ultimately create an operational “air traffic control” for hospitals — a centralized command and control capability that is truly predictive, learns continuously and uses advanced optimization algorithms and artificial intelligence to deliver prescriptive recommendations throughout the hospital system.
Photo: Pixtum, Getty Images
