ED The Cutting Edge
Use real-time data to reduce delays
From predicting the future to preventing it: The use of real-time data to monitor the ED system and intervene in real time
By James Espinosa, MD, FACEP
Overlook Hospital, Summit, NJ
I genuinely admire anyone crazy enough to stand up and be accountable for emergency systems in general, and emergency departments in particular.
We have responsibility for systems over which we have varying, and generally limited control. Management scholars may roll this sort of frustration into new terms like "stewardship," but at the end of the day, it still turns out that we need a better way of managing.
I would like to share with you a bit of a godsend of an approach that we have developed for use in the Overlook ED, and do a little future-think concerning the obvious implications of the strategy for hospitals and health systems. It has allowed us to propel ourselves from the 20th percentile nationally in patient satisfaction to the 99th percentile nationally. It has allowed us to do more with less resources. And it has allowed me, as an ED manager, some sense that we can prevent ED meltdowns.
Ray Bradbury was right. Ray Bradbury has been variously quoted as having said that the central theme of his brand of science fiction was not so much to predict the future as to prevent the future. He saw his work as a sort of interdiction on a future that was in some ways intuitively obvious to him. I share his intuition when it comes to ED management.
The more one knows about ED management, the more one realizes that the ED really is a sort of hub of a complex web of activities. One of the most frustrating intuitions in the world for an emergency physician to have, is to appreciate the incredible dependency of the ED on the rest of the hospital system.
The conventional approach. Over the years, many of us have become increasingly sophisticated at collecting, analyzing, and extracting the meaning of various ED data streams. We can graphically display the previous weeks, months, and years of many important variables. Some of us can use statistical software to make interesting inferences using the data, and make all manners of predictions about future behaviors of our systems.
Without a doubt, these are vital activities, and are core processes in our management strategies. Recent advances in tracking systems make the pool of potential data streams richer and deeper. Targets emerge for re-engineering efforts. In a recent edition of ED Management, we presented examples of efforts to re-engineer x-ray cycle times, for example.
One of the exciting aspects of the work we do at Overlook in this area is the use of "real-time data." We have created a "dashboard" or "instrument panel" of eight critical cycles. These cycles as displayed as small multiples on one screen of our ED tracking system. This is not a proprietary notion. We have shared the concept with everyone who wants to learn. It really is an adaptation from industry. We use a tracking system called "PaTrack," but any tracking system could be used this way.
I adapted the idea through researches into industrial quality management. It turns out there are, for example, steel mills that monitor, in real-time, several critical manufacturing parameters. These include steel thickness, heat, cooling bath temp, and so on. Advanced manufacturing plants and all sorts go a step further, however. Having re-engineered, they look at a core battery of processes in real-time, and make interventions and adjustments in real time, in order to influence the outcome of the batch of product at hand.
In our adaptation, we look at parameters critically linked to ED efficiency and to patient satisfaction. Data is displayed on one large chart, comprised of eight trend charts, on a 21-inch touch screen. This sort of single chart, with component charts, is known as a "small multiple" chart.
Customer-derived specification lines are laid in on the trend charts. The charts show the status of each function going back three hours until the present, in 15 minute bins of time displayed as bars or points on a line. The parameters we see can be changed on the fly, but we have settled on:
2. Occupancy in the department
3. Arrival to bed cycle time
4. Arrival to RN cycle time
5. RN contact to physician contact cycle time
6. Arrival to physician cycle time
7. X-ray cycle time
8. Admission cycle time (decision to admit, to time patient leaves department)
Example: Protecting the gains made in the x-ray re-engineering project. The first of the parameters developed was "x-ray cycle time." A run chart was developed to show x-ray turn cycle time performance in 15-minute bins of time. The radiology technicians and the ED staff monitor this vital sign, along with seven others, at all times.
The customer-specified goal of 30 minutes or less has been translated into a specification line on the "vital sign" indicator of 25 minutes. The technicians know that if performance exceeds the goals for three consecutive 15-minute periods, an intervention is needed. The intervention may include a temporary increase in capacity, through pulling a technician from another area of the hospital, or even that a technician may need to be called in from home. This approach has been very well received by the ED technicians.
The more one sees the system this way, the more one sees "systems pathophysiology" patterns emerging. We see three major sorts of patterns at present. Others will, no doubt, emerge.
• Downstream delay patterns. For example, delays in admission cycle times will manifest first as several 15-minute periods of cycle time trending upward. If the problem persists, and is of a serious enough nature, the next effect will be that ED occupancy will increase. The department becomes virtually smaller, and then it becomes difficult to bring patients into the department, leading to increases in all of the arrival-to-practitioner cycles. Ultimately, this can lead to terribly prolonged arrival-to-physician contact cycle times, which will lead to lower patient satisfaction scores. x-ray cycle times are also a downstream problem. Slow x-ray time can effect the entire ED efficiency in a similar manner.
• Upstream delay patterns. For example, the number of patient arrivals (demand) outstrips available staff (capacity). The first impact will be seen in the arrival to bed cycle, leading to arrival-to-practitioner cycles. Note that once again, in time, unabated, this can lead to prolonged arrival-to-physician cycle times.
• Midstream delay patterns. For example, demand on practitioner resources outstrips capacity. Given sufficient beds in the department, the immediate effect will be on bed-to-nurse, and bed-to-physician cycles, then leading to prolonged arrival-to-practitioner times.
Note that in every sort of pattern, physician contact cycle time is ultimately effected. Physicians may appear to be the root cause of every one of these syndromes. It is tempting for the casual observer to blame all of these scenarios on physicians.
What’s the point? Real-time adaptation of capacity to demand. For each of these scenarios and more, we have strategies in place. These are beyond the scope of this column, but are based on best practices and common sense. Using the x-ray cycle time as a paradigm, we first re-engineered ED radiology services, and brought that cycle from 70 minutes to 23 minutes. Then the question changed to getting the most out of the improved system.
The radiology techs look at their cycle on the screen (1 of the 8), and can "swing-in" capacity from the operating room (OR), or from outpatient services, based on the parameters displayed. (3 or more consecutive points trending about the customer spec line of 30 minutes). They release this temporary capacity when the system stabilizes. This sort of approach carries over for physician capacity (arrival to MD treatment time), ED nursing, and, most critically, for the cycle time for admitted patients to be transferred to the floors. The outcomes have been significant. We reduced arrival-to-physician evaluation time (critically linked to patient satisfaction) from 31 minutes (median) to 16 minutes (P = > 0.0005). X-ray cycle time dropped as above, with similar significance. Admission cycle time dropped to about 60 minutes. We recently won the 1998 Press-Ganey Award. In addition, our scores in the most recent quarter were at the 99th percentile.
Future think beyond the functional silo mentality: Why can’t entire hospitals adopt this approach? ED management of this sort is necessary but not sufficient. Anyone reading this column who works in the real world immediately recognizes that cycle times in the emergency department (ED) are dependent on a great many hospital-wide cultural, structural, and state-dependent factors. What is the status of telemetry beds? What is the critical care census? What is the hospital occupancy? It seems to us that the same sort of calculus would apply. What if hospital managers could see the behavior of critical hospital processes? Would patterns and trends emerge? One would think that much could be done to intervene at the hospital management level, in real-time.
The power of the strategy: Empowerment of staff to intervene on protocol. Certainly, hospital and ED managers do this sort of intervention every day, based on data available. Part of the power of real-time data is to empower others to make decisions. In the case of the x-ray redesign, agreements were inked in advance concerning exactly what sorts of escalating interventions could be made by the technician on duty.
Hospital systems linkage. Many groups of hospitals have formed alliances of various sorts. In most of these relationships, there is an interdependence based on levels of care provided and centers of expertise. From the ED perspective, we can see that in practice, the transfer patterns are often less efficient than would be desired. Hospitals could learn to visualize their statutes individually, and share these with their partners. Certainly, this would make tremendous sense around critical "product lines," such as critical care.
Other implications. Why is every flu season a surprise? Last year, ED and hospital-related influenza cases created heroic situations in many areas of the country. The media predicted the march of the virus across regions, and yet most hospitals had not geared up. In many of these hospitals, capacity lay fallow and dormant, awaiting additional staffing in order to open units. Contingency plans could have been in place, awaiting real-time data to activate them.
Conclusion: Not a panacea, just another tool. This notion of real-time data, real-time trending, and real-time algorithmic interventions is not a new invention, nor is it a panacea. It is another tool for us to deploy. For those of us in the ED who realize that they are managing a system with remarkable dependencies on the efficiencies and capacities of other hospital functions, it may well be an approach that makes the quality of life better not only for our patients, but also for ourselves as providers.