This article was originally published in the Spring 2002 issue (Volume 10, Number 3) of Quality Management in Health Care. It appears on the AGI Website with the permission of Aspen Publishers, Inc.
Applying the Theory of Constraints in
Health Care: Part 1 - The Philosophy
Anne M. Breen, Tracey Burton-Houle, and David C. Aron
The imperative to improve both technical and service quality while simultaneously reducing costs is quite clear. The Theory of Constraints (TOC) is an emerging philosophy that rests on two assumptions: (1) systems thinking and (2) if a constraint “is anything that limits a system from achieving higher performance versus its goal,” then every system must have at least one (and at most no more than a few) constraints or limiting factors. A constraint is neither good nor bad in itself. Rather, it just is. In fact, recognition of the existence of constraints represents an excellent opportunity for improvement because it allows one to focus ones efforts in the most productive area -identifying and managing the constraints. This is accomplished by using the five focusing steps of TOC: (1) identify the system’s constraint; (2) decide how to exploit it; (3) subordinate/synchronize everything else to the above decisions; (4) elevate the system’s constraint; and (5) if the constraint has shifted in the above steps, go back to step 1. Do not allow inertia to become the system’s constraint. TOC also refers to a series of tools termed “thinking processes” and the sequence in which they are used.
Key words: constraint, interdependencies, quality improvement, systems thinking
Introduction
The imperative to improve both technical and service quality while simultaneously reducing costs is quite clear. This imperative reflects the competitive pressures resulting from a number of inexorable forces including consumerism, an aging population, designer drugs, and technological innovation, among others. While health care has been relatively slow to adopt the management philosophies of quality improvement such as total quality management utilized in other sectors of the economy, the urgency has become even greater. A wide variety of philosophies, strategies, techniques, and tools have been tried with varying success. The Theory of Constraints (TOC) is an emerging philosophy that offers some distinct advantages, both theoretical and practical.1-4 The purpose of this article (Part 1 of a two-part series) is to introduce the basic concepts of TOC and illustrate their application to a “simple” problem.
TOC originally was developed by Eliyahu Goldratt, a physicist, and published in a novel format in The Goal.1 There are two primary assumptions of this approach. First, it is a form of systems thinking that looks at an enterprise as a complete and complex system where any number of constituent parts interact with one another. Second, if a constraint “is anything that limits a system from achieving higher performance versus its goal,” then every system must have at least one (and at most no more than a few) constraints or limiting factors. A constraint is neither good nor bad in itself. Rather, it just is. In fact, recognition of the existence of constraints represents an excellent opportunity for improvement because it allows one to focus one’s efforts in the most productive area -identifying and managing the constraints. TOC also refers to a series of tools termed “thinking processes” and the sequence in which they are used. These tools enable people to analyze their systems to determine what to change about the system, what to change the system to, and a way to cause the change. In Part 2 of this series (to appear subsequently), the authors will detail some of these tools.
A Simple System
TOC views every organization as a chain of interdependent events (or processes) where the performance of each event (or process) is dependent upon the previous event. Thus, the structure of a system and specifically the structure of a system’s interdependencies defines its performance. Consider a relatively simple system of a physician’s office or clinic. The steps in the process could be patients checking in, filling out forms, having vital signs taken by a nurse, seeing the physician, seeing the nurse for a prescribed procedure such as vaccination, and so forth. These steps could take place in a simple linear sequence or chain as shown in Figure 1. Each link in this chain has the ability to perform its tasks at different average rates. In this example, the first resource can process 13 patients, charts, or blood samples per an hour; the second can process 17, and so forth. One may think that this process can produce 13 per hour, the average of all resources. In fact, this process or chain only can produce an average of eight per hour. The chain is only as strong as its weakest link and the rate of the slowest resource in this example, the weakest link, is eight. This is true regardless of how fast each of the other resources can process individually or how much work is stuffed into the pipeline, how complex the process or set of interconnected processes is to complete. Moreover, improving the performance of any link besides the constraint does nothing to improve the system as a whole.
Figure 1. Top panel: A simple system depicted as a chain. Middle panel: Each of the links in the chain has a different capacity; throughput is determined by the capacity of the weakest link. The average capacity is 13; system throughput is 8. Lower panel: Impact of cutting "excess" capacity. Constraint (physician) must do some work previously done by other links in the chain, reducing throughput from 8 to 3.
Statistical Fluctuations and Dependent Events
The situation in the simple chain described above is actually worse, or at least less predictable, than it appears. As a patient proceeds through the system, there are many factors that can affect the work rate and introduce variation into the process. For example, some patients arrive on time while others arrive late. A new patient to the clinic may require more time to fill out forms than an established patient. The clerk may or may not be preoccupied with another task. Similarly, some patients require more time with the physician than others. If there were only one step in the process, the rate would average out over time. However, in a system of dependent processes, when variation or statistical fluctuation occurs at any step, it has an impact further down the chain. Time lags are generated, accumulate, and increase further down the chain and the performances of the system actually may be worse than the average capacity of the weakest link. The constraint may be forced to wait for work or be unable to pass its product on to the next link in the chain if the latter is busy. Whenever the constraint is idle and not performing productive activity, that time is lost to the whole system. Similarly, a patient lost to the constraint is a patient lost to the system. This is only the beginning of the problem.
Managing by Trying to Cut Costs: Effects of Traditional Cost Accounting Practices
Like most organizations, economic value is demonstrated using traditional cost accounting practices, with particular emphasis placed on cost per service. Improvement in cost per service usually is accomplished by driving down all costs while increasing all resources’ productivity and, in the case of the clinic or physician’s office, the number of patients that go through the system. As a result of everyone’s education, training, and/or experience, this approach appears to be logical. However, the decisions, measurements, and behaviors this approach typically drives can be devastating, yielding effects that not only increase costs overall but significantly decrease systemwide productivity and revenues. It is not uncommon for organizations to do things like mandate across-the-board staffing cuts. Across-the-board cuts do not differentiate between the weakest link and the stronger links. It would make more sense to decrease the capacity of stronger links in the chain, leaving the performance of the weakest link, and therefore the entire chain, intact. In most organizations, productivity improvements are sought from each and every link/resource/department in the chain. Improving the rate at which stronger links in the chain process work has no impact on the overall performance of the chain. Such practices consume resources with little or no impact on system-wide performance. It would make more sense to focus scarce resources on improving the weakest link and, thus, the performance of entire chain.
The impact of demanding that this chain see 10 patients per hour when it only has the capacity to see eight would be increased wait times, chronically late clinic closing times, dissatisfied customers and staff, and, frequently, significant decreases in the quality of care, to mention only a few. It makes more sense to match the capacity of the weakest link to the demands placed upon the chain than vice versa. In fact, rather than optimizing the system by attempting to make each step most efficient, the system should not be balanced by equalizing the capacity of every step by trimming “excess capacity” at the steps in the process that are not the constraint. Rather, the system should be balanced so that the constraint is working at its best. This means that there must be excess capacity at the non-constraints. The non-constraint processes should be utilized to ensure that the constraint works smoothly. Therefore, they should not stay busy all the time. If a system is doing well, not more than one of its component parts will be. If all links in the chain are performing well, the system will not be. To operate within this conceptual framework requires synchronization across the links of the chain, i.e. resources, departments, and so forth as well as a new set of performance measurements.
Performance Measurements
The only true measurement of success (or failure) of any system is how well that system performs relative to its goal. In other words, success or failure is dictated by how many “units of the goal” (throughput or T) that system generates in a given period of time. In a for-profit enterprise, one assumes that the goal of an organization is to make money both now and in the future. Consequently, the units of the goal are clear. The three operational performance measurements are throughput, inventory, and operating expense, but these are defined somewhat differently in TOC than in standard cost-accounting.2 Throughput (T) is the rate at which the system generates money through sales and is represented as sales minus “totally variable” cost; output that is not sold is not throughput, it is inventory. In a for-profit medical practice, it would represent all the money flowing into the practice through the variety of services provided to patients minus the totally variable costs. Inventory (I) is all the money invested in things the system intends to sell. Inventory includes any physical inventories such as raw material, work in process, unsold finished products, and includes tools, building, and equipment. (Conceptually, patients waiting for treatment can be viewed as raw material or work-in-progress inventories.) Operating expense (OE) is all the money the system spends in turning inventory into throughput. In other words, OE includes all the money going out of the practice such as wages, salaries, utility expenses, and interest payments.
In the not-for-profit sector, the goal is somewhat different. For a health system, the goal could be defined as “to provide quality health care to a particular population now and in the future” and the operational measurements must be different. Throughput is “units of health generated” (not an easy thing to measure, to be sure). Moreover, this measure of T alone, however, doesn’t give the whole picture. One also would want to know how effectively the organization utilizes the one resource it’s provided – money - to maximize T. There are two ways to utilize money in the organization: (1) expenditure on current operations or (2) investment (or savings) in future operations. Therefore, one would want to measure how much T was generated per dollar invested in the system (T/$I) and how much T was generated per dollar spent in operating expense (T/$OE). In industry, these bottom-line measurements are called turns and productivity, respectively -only their T is dollars of profit, their “unit of goal.” These bottom-line measurements tell what a system’s overall performance was relative to its goal in the past. Because one also is interested in a system’s future performance, there are other indicators that one must look at as well. Minimally, these indicators should include what are called necessary conditions - defined as those things that, if violated, would jeopardize the very goal, and future, of the system. In a health care organization, these necessary conditions may include (but not be limited to) patient satisfaction, staff satisfaction, compliance with external requirements (e.g., accreditation boards, regulatory bodies), and operating within budget. Indicators that track these necessary conditions are important and must be considered alongside the bottom line measurements in any decision-making process; jeopardizing these indicators jeopardizes the future. Thus, one needs indicators that measure these necessary conditions. The strategies of the organization determine what these measurements should be. Improvement in the organization means progress toward the goal as measured in goal units. After the boundaries of the system to be improved are established - e.g., a single clinic or all the clinics in an organization - one then can utilize the steps in the TOC quality improvement process.
The Steps in the TOC Quality Improvement Process
The fundamental principle of TOC provides a focus for a continuous improvement process and has been reviewed in detail.1-3 The five focusing steps in this process are shown in Table 1. They are the following:
- Identify the system’s constraint(s): A physical constraint can be readily imagined, e.g., the number of examination rooms, the physicians, nurses, or clerks. In many health care organizations today, the constraint is the support staff. When it comes time to cut costs, most organizations prefer to cut clerical staff before physicians (who, in general, should be the constraint). The result is that providers spend more time doing clerical work and less time seeing patients: T goes down, T/I goes down, and T/OE goes down - all in the wrong direction. Policy constraints, such as ineffective policies, measures, or behavior patterns, are usually far more damaging to the system but are more difficult to identify and eliminate. In general, one would like the constraint to be the resource for which it is most difficult or expensive to create capacity. In a clinic, this would be the physician.
- Decide how to exploit the system’s constraint(s): If the constraint is physical, the objective is to make the constraint as effective as possible. Just as the weakest link of a chain dictates the strength of the entire chain, the constraint dictates the maximum T of a process. A moment lost on the constraint is a moment of T lost for the entire process. So, after the constraint is identified, there must be a decision on how to squeeze the most T out of it (without, of course, jeopardizing any necessary conditions like staff satisfaction or compliance to budget). This can take many forms, depending upon what the constraint is. If it is exam rooms, the question may be how the clinic uses them to maximize the amount of T that can be generated through them. If it is providers, how can the clinic delegate tasks previously done by physicians to others so that physicians can focus on seeing patients and generating T? A policy constraint should not be exploited but rather be eliminated and replaced with new policies, procedures, measurements, and so forth to drive the desired behaviors needed to maximally exploit the constraint. The clinic may choose to utilize its resources very differently than in the past. Each time such a decision is made, the clinic must be attentive as it may shift the location of the constraint. If the constraint’s location shifts, the clinic may find itself continuing to exploit what is no longer a constraint but a non-constraint. When that happens, changes do not yield direct, if any, bottom line improvement.
- Subordinate/synchronize everything else to the above decisions: Subordinating or synchronizing everything else to the above decisions is the most difficult yet most important step. Without subordination, the plans to exploit the constraint will not occur or there will be difficulty in effectively converting the output of the constraint into maximum T. In either case, T is lost for good. After a clinic/office decides how to exploit its constraints, it must develop strategies for subordination/synchronization: how all other resources - all non-constraint resources - will operate to ensure that the plans to exploit the constraint and effectively convert its output into maximum T are executed. This includes how patient are scheduled to arrive at the clinics, how check-in is done, how exam rooms might be used, how consults are scheduled - every step in the process needs to be subordinated and synchronized as must every policy, measurement, decision, plan, and so forth. In addition, a key part of subordination and integration is identifying where buffers should be placed to protect the system’s T from variation (i.e., Murphy’s Law, statistical fluctuations, etc.) and how those buffers should be managed. A buffer is inventory or time that protects the physical constraint from running out of work. Using these buffers and an “early warning system” that indicates when intervention is needed to protect against lost T before it occurs fosters proactive rather than reactive decision making and response. It is important to bear in mind that shifts in the constraint have an impact on this step. Each time the constraint shifts, the policies, measurements, procedures, and so forth that were put into place to exploit the constraint and subordinate everything else to it must be reevaluated.
- Elevate the system’s constraint(s): In contrast to step 2 in which attempts are made to increase throughput without spending significant amounts of money, this step requires investment in the constraint. For example, it might mean hiring another physician or expanding the physical plant in order to increase capacity. The decision to do this or not is a strategic decision.
- If in the above steps the constraint has shifted, go back to step 1: Do not allow inertia to become the system’s constraint. In addition, an optimal solution tends to deteriorate over time as the environment changes. TOC is a process of ongoing continuous improvement.
These five steps of TOC provide the foundation for all of TOC’s generic solutions to a wide variety of problems including not only management of processes, inventory, and supply chains, but also project management and decision making.
Examples
The published data on use of theory of constraints in health care are limited to a few single institution case studies.5-8 For example, Womack and Flowers reported the experience of the 366th Medical Group, a U.S. Air Force Base providing inpatient and outpatient services.6 The target was to meet a 7-day standard for a routine primary care appointment. Following the intervention, the waiting time fell from an average of approximately 17 days (range 9-24) to an average of approximately 4.5 (range 2-7). This was accomplished without additional cost. Moreover, they demonstrated that by elevating the constraint at a cost of less than $200,000, they could increase their capacity to enroll an additional 800 covered lives with projected revenues of $1.6 million. Staff at the Radcliff Infirmary in Oxford, England, utilized TOC to improve waiting times for neurosurgery and ophthalmology.7,8 They noted a 100 percent reduction in elective cancellations as well as increases in throughput of 16 percent and 20 percent in neurosurgery and ophthalmology, respectively, without any additional resources or major re-engineering project.
Conclusion
TOC offers a logical and rigorous method for analyzing and improving the performance of a health care enterprise. Part of its power rests in the fact that much of it is “common sense.” It has been applied in a number of areas in health care both in the U.S. and England although little has been published. In part, this reflects its novelty. In Part 2 of this article, to be published subsequently, the authors address some of the tools that facilitate the identification of constraints and their management. These tools, termed the thinking processes, reflect the application of the thinking processes of the hard sciences - rigorous cause and effect - to understand and improve systems of all types, but particularly organizations.
References
E.M. Goldratt and J. Cox. The Goal: A Process of Ongoing Improvement, 2d rev. ed. Great Barrington, MA: North River Press Publishing Co., 1992.
H.W. Dettmer. Breaking the Constraints to World Class Performance. Milwaukee, WI: ASQ Quality Press, 1998.
D. Lepore and O. Cohen. Deming and Goldratt, The Theory of Constraints and the System of Profound Knowledge. Great Barrington, MA: North River Press Publishing Co., 1999.
H. Roybal, S.J. Baxendale, and M. Gupta. “Using Activity-Based Costing and Theory of Constraints to Guide Continuous Improvement in Managed Care.” Managed Care Quarterly 7 (1999): 1-10.
R. Kershaw. “Using TOC to “Cure” Healthcare Problems.” Accounting Management Quarterly (Spring 2000): http://www.manag.com/spring00/sp00kershaw.htm. Accessed 22 January 2002.
D. Womack and S. Flowers. “Improving System Performance: A Case Study in the Application of the Theory of Constraints.” Journal of HealthCare Management 44 (September/October 1999): 397-405.
NHS Executive Action on Cataracts. Good Practice Guidance (February 2000) http://www.doh.gov.uk/pub/docs/doh/cataract2.pdf. Accessed 22 January 2002.
NHS Learning Network. Focus on Ophthalmology Waiting Lists. ImpAct (2000) http://www.jr2.ox.ac.uk/bandolier/painres/download/ImpAct6.pdf. Accessed 22 January 2002.
Copyright © 2002 Aspen Publishers, Inc.
