Rx for the ER

People seeking emergency treatment has soared over the last decade, creating an overcrowding crisis in emergency departments.

By Joseph Guarisco and Douglas A. Samuelson

Healthcare is being challenged operationally more than ever before. From that perspective, it may be the last of the “mom and pop” industries. Surely, as an industry, it exhibits technologic excellence in terms of its diagnostic and therapeutic innovations. However, the drivers that force other industries to excel and continually improve in another area of importance – service delivery – have been absent in healthcare. Healthcare in the modern era has remained in select ways a noncompetitive industrial sector, but that is about to change.

Patient choice, ripened by increasingly available hospital performance data, has moved to the forefront and reflects largely patient satisfaction with clinical services. Customer satisfaction is a major driver of any service delivery system. The economic incentives to develop and sustain service delivery models that are viewed by the patient as efficient, useful and valuable have been to a large degree nonexistent in a hospital environment. Additionally, hospitals operate on very thin financial margins, creating risk avoidance and therefore reluctance to invest in operations that potentially deliver more efficient and productive healthcare services [6].

Now, however, hospital performance data in myriad clinical and service areas are increasingly entering the public domain and forcing rapid change. So as payers move to performance-based reimbursement and as patients increasingly demand relevant information, more choice and better services, healthcare in the United States will move to embrace the same operations management science that has yielded a competitive edge in many other sectors of American industry [5].

As an emergency physician for 30 years, the first author of this article can speak to one example: the historical failure of hospitals to develop a responsive service delivery model for emergency services. Emergency rooms, now widely known as emergency departments (EDs), are perfect examples of a service delivery system with seemingly random demand and for which failure to match that demand carries significant clinical risk to the patient, as well as financial risk to the hospital. Emergency departments typically function in an environment in which the demand for services (beds and providers) exceeds the supply at peak hours if not the majority of the day. It’s the classic patient (as customer) nightmare: waiting. It’s more than just inconvenience; sometimes waits can impact patient morbidity and even result in a patient’s death [9].

The Impetus for Change

What change management authority John Kotter calls “the burning platform for change” began at Ochsner Hospital Center in New Orleans in 2005 a few months after Hurricane Katrina [7]. The storm rendered 70 percent of the healthcare services in New Orleans partially or totally unusable. As the population returned, ED volumes ramped up seemingly overnight to as much as 180 percent of pre-Katrina baseline averages [13].

Figure 1 shows visits per day for one year past Katrina, as a percentage of the pre-Katrina baseline. Figure 2 displays the average minutes waiting to be seen, by day, up to one year post Katrina. Twenty minutes was the pre-Katrina benchmark average. Random variation produced much larger peak waits on bad days.

Figure 1: ED visits per day at Ochsner Hospital as a percentage of the pre-Katrina baseline.

Figure 2: Average ED waiting times by day at Ochsner post Katrina.

Clearly, Ochsner needed to find another way to operate its ED. The hospital’s response was to innovate and develop a new, more efficient service delivery model built around operations research and management principles, while soliciting insights from medical practice to identify key bottlenecks in the ED’s processes and procedures. The model, called qTrack, ensures that removing one bottleneck does not simply shift the backup to another bottleneck. The result: a dramatic reduction in patient wait times and in the percentage of patients leaving without being seen.

Hospital management had several good reasons to invest in a more efficient service delivery system, including:

• Healthcare is in fact transitioning in a competitive environment and patient satisfaction is critical to success, as patients (customers) express choice based on service quality. The ED is the front door of most hospitals and is the conduit for 50 percent to 75 percent of hospital admissions. Patients’ experiences there are an important component of their satisfaction with the hospital [12].
• Both clinical quality and patient safety are impacted by service delivery. Most surrogate quality metrics in emergency medicine are time-critical measurements, and clinical outcomes are affected by delays such as delays in delivery of clot busters for patients with strokes and heart attacks or delays in the delivery of antibiotics for patients with pneumonia. The Ochsner ED is ranked in the top 5 percent of EDs nationally by HealthGrades, the only ED in the country to earn this ranking two years in a row [10]. So in many ways, the reengineered workflow developed at Ochsner enables some of that success.
• Hospital growth and profitability are both affected when patients choose not to wait and leave prior to evaluation. With ED volume of 50,000 patients a year, the annual loss of revenue is estimated to be $500,000 net for every 1 percent of patients who choose to leave prior to evaluation. It is not unusual for hospitals to have up to 7 percent of arriving patients leave before they are evaluated due to delays in their care [14].

Emergency departments are dynamic systems and function in an environment with extreme demand variability. As demand changes and as variation in that demand increases, so too must the available capacity. The response to service demand must occur proactively and predictably.

The OR/MS Solution

The Ochsner staff and management see the new service delivery model as having “cracked the code” in addressing the chronic inability of hospitals to deliver emergency services on demand in an environment with high variability. This service delivery model succeeds fundamentally by creating virtual capacity for its first most constrained resource, the bed. The ED bed is the major and most expensive resource that most EDs in the country run out of every day. The delivery model has other key benefits too, such as yielding cost reductions and productivity enhancements that provide the financial returns to fund increases of its second most constrained resource, the provider thereby reducing over utilization. The impact of excessive utilization of resources on wait times is well known, increasing exponentially as utilization approaches capacity. Typical EDs operate at 90 percent or more of capacity, leaving little room to handle surges in demand.

The OR/MS solution involves multiple initiatives: 1) creation of both additional real and virtual capacity; 2) acquisition of data (analytics) to better understand, predict and resource demand and variability with much higher probability so as to make best use of that new capacity (otherwise that new capacity is wasted); and 3) implementation of a provider staffing model that cost-effectively matches resources (beds and providers) to predicted demand.

How the Model Works

Emergency departments are complex operational arenas managing interdependent processes. The standard workflow of a typical ED comprises several key steps: registration, triage, bed assignment and provider evaluation (Figure 3). Many other downstream processes are important, such as imaging and laboratory turnaround, but the most critical patient care delays occur in waiting for that initial provider evaluation.

Figure 3: ED work flow stream. Source: Joe Guarisco, Ochsner Health System

Typically both registration and triage in most emergency rooms are each a laborious 10-minute process. In the new service delivery workflow model, both processes are lean and reduced to a one-minute interaction, an 80 percent time savings. But this change alone would merely pass the patient on to another bottleneck, to no advantage. As previously stated, the major benefit is in addressing the first of two critical resources that present as a bottleneck: the bed.

As simple as it may sound, the process embodies a set of triage rules using a scoring system ESI (Emergency Severity Index) developed by Dr. Dave Eitel based on both acuity and expected resource utilization to determine which patients need a bed and which patients do not need a bed [3]. A key understanding is that not every patient needs a bed. It is somewhat surprising that EDs did not think long ago to segment patients into two groups based on “bed needs” in managing this scarcest resource. The typical patient arrival pattern at peak times (if not most of the day) creates a census commonly twice as large as the bed capacity. One can then understand why patients wait and why this kind of model conserves the bed resource works to create capacity. Figure 4 shows the Ochsner Main Campus ED with a bed capacity of 40 and a census of 70 on a given day. The graph displays the capacity problem typical of most EDs.

Figure 4: ED patient census vs. hour of day. Source: Joe Guarisco, Ochsner Health System

Figure 5: Visio Workflow qTrack - Source: Joe Guarisco, Ochsner Health System

To look at it another way, typically 75 percent of patients are discharged from an emergency department, further validating that not every patient needs a bed during their stay. In the Ochsner ED, as in other systems that used this segmentation (split flow) concept, 60 percent of patients never utilize (consume) a bed. This was clearly shown in the excellent work done by Banner Health System in its grant work on Door to Doc [2]. These patients are managed in a fashion typical of most ambulatory environments. Simply, the model generates capacity by virtue of this process.

Following registration and triage in this reengineered workflow, patients are next evaluated in “intake” rooms. Following the initial evaluation, these patients are then moved not to acute care beds but to continuing care areas or internal waiting rooms awaiting results of tests. The analytics must determine just how many intake beds will be needed to meet incoming demand with 90 percent probability (a triage super highway) [2].

So how many intake beds does one need? Figure 6 from the Banner Health System shows arrival data for seven different hospitals. The patterns are remarkably similar by facility, by day of week, by hour of day, etc. and are therefore predictable. It is a perception that emergency department demand is so unpredictable one cannot possibly begin to plan on successfully addressing this from an operational management and resource perspective [2].

Figure 6: Hourly patient arrival distributions for seven hospitals. - Source: Kevin Roche, Banner Health System

Many hospitals manage the emergency department without any knowledge of this fairly common demand curve, and those that do typically manage to the average demand without any respect for variability [1]. If one manages resources to average demand, half the patients will be unhappy while experiencing long waits for service. To understand this better, consider Figure 7 that shows a box plot with a distribution of patient arrival rates by hour of the day. The black line demonstrates the average rates with the top data whisker “†” indicating 90 percent probability for arrival rates [8].

Figure 7: Box plot (ED pts/hr arrivals vs. hour of day). - Source: Ed Popovich, executive vice president, MX.com

Many will argue that the demand variation in emergency medicine is somehow different from that in other industries such as restaurants, grocery stores, factory supply houses or any organization or industry involved in service delivery [4]. The reality, however, is that emergency department demand and variability are measurable and understandable and, therefore, within the science of probability, predictable [11]. No, we cannot precisely predict demand in a highly variable area such as an emergency department, but, as in other industrial forecasting; one can predict the probability of demand being within certain limits on any given day at any given moment. Providing sufficient provider resources to meet the 90-percent patient arrival probabilities is the second constraint. In addition to sufficient intake beds, one must provide sufficient providers to match incoming patient arrivals. So let’s now discuss why and how bad margins work against hospitals in solving this math problem.

Financial Impact

Successful service delivery industries and organizations create a staffing cost structure that allows them to resource variable demand sufficiently. The cost of service vs. cost of waiting is best visualized in the graph depicted in Figure 8. It’s the source of the great debate every hospital administrator struggles with, and a choice is made where along that continuum one lands based on affordability.

Figure 8: Cost of service vs. cost of waiting. - Source: Kevin Roche, Banner Health System.

The cost of service is straightforward. The cost of waiting is shown in the cost analysis depicted in Figure 9 where the net loss revenue based on 1 percent elopement rate for patients that get tired of waiting and leave.

The segmentation of patient streams into complex, high-acuity disease states and lower acuity, less complex disease states is very important. This segmentation allows the creation of a “higher margined” staffing structure for the lower acuity stream. Within the lower acuity patient stream, 80 percent of the provider work does not mandate a highly skilled and highly valuable emergency physician.

Most of the work required of a provider involves gathering physical supplies, collecting data, constructing electronic lists, verifying allergies, reconciling patient medications, computer entry of the encounter details and creating electronic prescriptions and discharge instruction. These are all tasks that can be done as fast if not faster by physician extenders. Advanced practice clinicians (mid-level providers such as physicians’ assistants and nurse practitioners) at 25 percent of the cost of an emergency physician can deliver (with supervision) excellent patient care in this setting. This allows staffing at double the hourly productivity (half the cost of the traditional staffing model) increasing the probability of meeting the upper limits of variable demand and tolerating the lower limits of variability with minimal waste (cost). Good customer service becomes affordable.

The impact on two key metrics is shown in Figure 10. The v1 marker shows the success of the initial implementation, which was followed by significant ED volume growth that eroded the early success. This growth required tweaking the model quantitatively, resulting in continued improvement as shown by the v2 marker. Demand matching is a continuous process. The improvement from the approach is clear, though, even taking into account the need for additional tweaking.

Figure 10: Impact on two key metrics.

Conclusion

The workflow model described in this article and other similar service delivery models have driven D2D (door to doctor) to its theoretical lowest limits [11]. Improving provider service delivery times (D2D) is the key driver in improving patient satisfaction, reducing risk and improving patient care quality and safety, thereby providing for better financials and competitive growth in the market place.

Joseph S. Guarisco, M.D. (Jguarisco@ochsner.org), is chairman of the Department of Emergency Medicine and system chief of Emergency Services for the Ochsner Health System in Louisiana. The Ochsner Health System manages seven emergency departments (ED) which combined have more than 250,000 visits a year. Board certified in emergency medicine, Dr. Guarisco also holds a bachelor’s degree in engineering and has significant experience in ED informatics and automation pertinent to the design, development and implementation of ED information systems. He has pioneered ED workflow redesign through innovative adaptation of queuing theory and engineering principles to ED patient flow processes. He was twice awarded Innovator of the Year in New Orleans for his work on qTrack© and the implementation of Web-based ED wait times for the Ochsner Health System.

Douglas A. Samuelson, D.Sc. (samuelsondoug@yahoo.com), is president and chief scientist of InfoLogix, Inc., a consulting and R&D firm in Annandale, Va. He has applied operations research to a wide variety of problems, with health care as one of his major interests. He is a frequent contributor to OR/MS Today and Analytics magazines.

References

1. Eitel, David, Silver, John, “Optimizing Emergency Department Throughput,” CRC Press. 2010.
2. Burdick, T.L., Cochran, J.K., et al., “Door to Doc (D2D) Patient Safety Toolkit,” Rockville, Md., Agency for Healthcare Research and Quality, http://bannerhealthinnovations.org/.
3. “Emergency Severity Index, Version 4: Implementation Handbook,” http://www.ahrq.gov/research/esi/esihandbk.pdf.
4. Eitel, David, “Grocery Store Math in the Emergency Department,” http://nowaited.com/grocerystore.htm.
5. Kenen, Joanne, “Hospitals Try New Approaches to Ease ER Crowding,” http://www.kaiserhealthnews.org/Stories/2011/January/14/Emergency-department-crowding.aspx?p=1.
6. Institute of Medicine (IOM), “Hospital Based Emergency Care: At the Breaking Point,” Washington, D.C., Institute of Medicine, 2006.
7. Kotter, John P., “Leading Change,” Harvard Business Press, 1996.
8. Ed Popovich, MX.com.
9. “Patient Dies in Hospital Waiting Room,” MSNBC
   http://www.msnbc.msn.com/id/19375461/ns/health-health_care/t/hospital-may-lose-license-after-er-lobby-death/.
10. HealthGrades, http://www.ochsner.org/lp/healthgrades.
11. Eitel, David, Samuelson, Douglas A., “O.R. in the ER,” OR/MS Today, August 2011.
12. Guarisco, J., Bavin, S. (Press Ganey Associates), “Validating the Primary Provider Theory in Emergency Medicine,” Leadership in Health Service, Vol. 21, No. 2, 2008 United Kingdom.
13. Guarisco, J., “Re-engineering ED Workflow: How a New Orleans ED Made a Comeback Using qTrack, Society of Health Systems,” Orlando, Fla., February 2008.
14. Guarisco, J., “Grand Rounds Presentation, Controlling Chaos: Faster Better Less Costly Health Care. How Would Toyota Do It?” Ochsner, New Orleans, La., June 28, 2011.