OR/MS Today - April 2002|
Health Care Management
Diagnosis: Mismanagement of Resources
Ailing health care system desperately needs a dose of operations research, so why aren't more OR professionals responding to the emergency
by Michael Carter
Health care is the No. 1 domestic industry in the United States and one of the largest industries in the developed world. Health care systems present many complex problems that could benefit from operations research-type analysis and applications. OR professionals, however, have generally neglected the field.
Less than 2 percent of INFORMS membersabout 180belong to the Health Application Section. Eighty of those are academics and another 30 are students. Twenty-two call themselves practitioners, 10 are consultants, and only two actually work at a hospital. Many OR people have written at least one paper on health care or have been involved in at least one health care projectand then moved on to something else. Very few OR people specialize in health care.
Why don't more OR people work in health care? The health care industry faces many of the same issues confronting other industries, but with some significant political differences. If you try working in the health care area without understanding the politics behind the issues, you are asking for some grief. At the same time, health care represents a huge segment of the economy, and it needs our help.
How Large is the Health Care Industry?
In 1999, the United States spent $1.2 trillion on health care ($662 billion in private money; $549 billion in public money, primarily federal Medicare and Medicaid funding), according to the U.S. Health Care Financing Administration. The $1.2 trillion represents 13 percent of the U.S. gross domestic product for 1999. (The percent of GDP devoted to health care in the United States has been dropping recently due in part to strong GDP in the past eight years.) In more personal terms, the per-capita annual expense for 1999 equates to $4,358.
Typically, it takes two to three years to collect reliable information on health care spending. The system is large and not well integrated, and reporting is slow. Hence, the figures for 1999 are the most recent real numbers. However, according to HCFA projections, the United States spent $1.4 trillion in 2001, which equals 13.4 percent of GDP or $5,043 per person. The HCFA's projection for 2008 is $2.3 trillion or 15.5 percent of the GDP.
Corresponding figures for Canada indicate that Canadian health expenditures totaled $56 billion in 1999, and accounted for 9.2 percent of the GDP or $1,828 per capita. The figures are from the Canadian Institute for Health Information, which estimates that 2001 expenditures rose to $64.2 billion, or 9.4 percent of the GDP and $2,066 per capita. (Note: All monetary figures are expressed in U.S. dollars and were reported as of December 2001 on the HCFA's and CIMI's respective Web sites. The conversion rate used: $1US = $0.62633CN.)
Most people are astounded by the per-capita expenditures. Individual health care costs are generally much lower than that, perhaps a few hundred dollars per year. In fact, the elderly consume the bulk of health care dollars during the last few years of their lives. My father lived to be 91 and was pretty healthy, but he was probably consuming more than $1,000 worth of medications per week by the time he turned 90. This is frightening news for a baby-boomer population that will be retiring during the next 20 years just as their health care demands are escalating.
Figure 1 provides a breakdown of U.S. health care funds in 1999. Public expenditures accounted for 45.3 percent of total spending; the remaining 54.7 percent came from private sources. Figure 2 shows where money spent on health care went in 1999. Other spending includes dental, other professional, non-prescription drugs, home care, research, public health and construction.
Figure 1: Source of U.S. health care funds in 1999.
Figure 2: Where the U.S. health care funds were spent in 1999.
The United States spends more than double per person on health care compared to Canada, yet several scholars claim that Canadians, on average, enjoy better health. Canadians live about three years longer on average than Americans, they have more surgical procedures per capita, but they also have to wait for non-urgent service.
Of course, averages are deceptive. There is no doubt that Americans have the best health care system in the worldfor those who can afford it. However, I do not want to get into a discussion comparing the relative merits of Canadian vs. U.S. health care systems. I only mention this to suggest that, perhaps, Americans are not getting full value for their investment. This is not just an American problem. My experience with the Canadian system leads me to believe that there is widespread inefficiency in Canada as well.
The people working in the system are generally dedicated to providing the best possible service. The problem is, the workforce and, more importantly, management, do not have the training or knowledge to make the best use of the available resources. In my opinion, no private industry would survive with the level of waste and inefficiency commonly seen in health care. I am convinced, based on years of experience in health care, that if I had an army of OR/MS professionals at my disposal, we could easily cut the cost of health care by 10 percent to 20 percent, and vastly improve the quality of the system in the process.
What about the rest of the world? The OECD reports on health expenditures for most developed countries. Table 1 provides a few of the figures for 1998. The United States spends far more on health care as a percentage of the GDP than any other developed country. Most developed countries, with the exception of Mexico on the low end and the United States on the high end, spend between 6.7 percent and 10.6 percent of the GDP on health care.
There are many reasons why OR/MS has not been more successful in the health care field, starting with a lack of interest and support on the part of health care managers. I doubt they appreciate what we have to offer, and that is probably our fault. "Introduction to OR/MS" is not a common component of most Masters of Health Administration programs.
Secondly, there is an attitude that "we" are diverting funds away from direct clinical care and disease research. It is difficult to convince managers, particularly in the Canadian system, to divert limited funds away from direct patient care into better administration.
Another major issue is the dreadful state of health care information systems and data. For example, it is universally acknowledged that there is no good way to determine the effectiveness of any health care program or treatment, since we don't have good tools to measure "health," and no information systems to record a person's "health" over time, even if we could measure it. Designing linear programming to maximize "health" is pretty difficult when you don't know what "health" means and have no way to determine the impact that expenditures have on "health."
In the language of LP, we know what the constraints are, but the objective function is still a mystery. (Refer to Doug Samuelson's 1995 article in OR/MS Today for a detailed discussion.) There are also serious research funding problems. Health care management research is not usually viewed as a core research area by engineering, medical or social science funding agencies.
When Worlds Collide
Health care is a business like no other. It has multiple decision-makers with conflicting goals and objectives. Glouberman and Mintzberg  have produced a clever framework to help explain why the health care system in all countries has proven to be virtually impossible to manage effectively.
First, consider the acute care hospital. Most hospitals in the United States, and virtually all hospitals in Canada, are not-for-profit, independent corporations. Glouberman and Mintzberg identify four different management groups (called four worlds) within the hospital as illustrated in Figure 3. Doctors and nurses manage "down" into the clinical operations because of their focus on patient care. Managers and trustees manage "up", toward those who control or fund the institution. Moreover, employees (managers and nurses) practice some management "in" the institution, while doctors and trustees manage "out" of the hospital, since they are technically not employees and are thus independent of its formal authority.
Figure 3: Four worlds of the general hospital.
Figure 4: Four worlds of society at large.
According to this scheme, the bottom left quadrant is the world of cure, which is characterized by short, intensive and (mostly) non-personal medical interventions. North American doctors typically do not work for the hospitals. They are private entrepreneurs who have admission privileges at a hospital. (Some doctors are salaried hospital employees, but the majority of doctors work on a fee-for-service basis.) To maximize their income, doctors make brief appearances when the patient needs a cure and intervention (treatment), and then they move on.
The bottom right quadrant represents the world of care. This is the world represented by nurses, the providers who work directly for the hospital on salary and typically account for the largest component of its operating budget. They work in their own internal management hierarchy and have a unique relationship with patients. They are the only providers who actually touch patients.
Managers represent control. They are employed by the hospital and are normally removed from direct involvement in clinical operations, but are responsible for its control. Since managers lack the knowledge to understand clinical operations, they control what they can, i.e., costs.
The world of community, formally represented by the trustees or Board of Directors, is often composed of community members. The Board is responsible for setting hospital policy and appointing senior managers. However, they are the people who generally know the least about health care or its delivery, since they neither work for the institution nor do they provide clinical services.
In this fractured environment, doctors and nurses form what is called the "Clinical Coalition." They form a coalition, based on the objective of delivering patient care, usually as a common front as patient advocates against managers and trustees. The nurses and managers make up the "Insider Coalition," since they are the ones who actually work for the hospital and have concerns about the day-to-day operation of the organization. They form a coalition against the outsiders (doctors and trustees) to preserve their hospital and their jobs. The managers and the Board of Directors form the "Containment Coalition." They form a coalition on the basis of strong concerns about budget constraints. Finally, the Board and the doctors make up the "Status Coalition." They share the prestige of being independent of the institution, and yet, they are at the top of its pecking order.
Unlike any private sector business, no one is really in charge of a hospital. Managers make resource allocation decisions, but doctors decide what the hospital does with those resources. A horizontal cleavage divides the clinical workers from the containment sector, and there is little cooperation between the two. Glouberman and Mintzberg have found that both doctors and managers tend to turn to the nurses for coordination and conflict resolution. Nurses become the hospital managers. This puts them in an awkward situation, since they do not have the authority to truly manage.
The same template can be applied to the overall social health system. In this case, the acute care hospital itself represents cure. Patients go to the hospital when they are really sick, and then get quickly discharged back to the community (home care, family doctor) where they receive basic long-term care. The hospital is somewhat beyond direct public control and thus "out" of the day-to-day community. Government agencies or insurance companies provide control. They are removed from direct care, yet they are responsible for funding. Finally, politicians and advocacy groups, like trustees in the hospital model, try to influence the system without being directly involved in funding or care.
Glouberman and Mintzberg argue that these four worlds, in both models, operate independently and without much cooperation. The frequent reorganizations common in health care at both the hospital and system level usually just affect one of the worlds. Unless these worlds become integrated, costs will continue to spiral out of control.
In health care situations, we typically want to minimize cost or maximize quality or, more likely, some combination of these two. On the surface, this sounds pretty straightforward, but if we look closer, even the definition of these terms"cost" and "quality"is open to interpretation. Cost to whom? The hospital? The government? The patient? The doctor? Whose cost are we minimizing? Do we want to minimize the cost per hospital visit (minimize care and length of stay) or do we want to minimize the overall annual cost? In the latter case, we should spend more on prevention, as more tests now may mean avoiding a much more expensive episode later. When we minimize hospital costs, we often simply transfer those costs from the hospital to the family who must provide support or hire home-care nursing. In other cases, we want government to spend money on prevention programs that save social costs later, but these costs may not translate into government savings, which makes them difficult to cost justify.
One hospital administrator I met facetiously pointed out, "We all die eventually. The longer we keep you alive, the higher the cost to society. Smokers are cheap. They make tremendous contributions in tobacco taxes, and then typically they get lung cancer in their mid-sixties and die pretty quickly before they have a chance to get truly expensive. Therefore, to minimize cost to society, we should encourage people to smoke." This silly example illustrates the dangers of putting too much emphasis on health care cost, and not enough emphasis on quality.
But how do we maximize quality? Qualityreal patient outcomesis hard to measure. Once you leave the hospital, you're gone. Most hospitals have no information on what happened to you once you walk out the door, unless outcomes are so bad that you are re-admitted within a short period due to complications.
Doctors are the gatekeepers, but they don't care about cost. They are required to be the patient's advocate. Personally, I don't have the expertise to know what I need as a patient. I rely on my doctor to decide if I need more tests. And if it were me, I would say do the tests, the heck with the cost, especially if I don't have to pay for them.
Maximizing quality is also quite ambiguous. Do you maximize the quality of the outcome for a particular episode of care? Or do you try to maximize the patient's quality of life?
One particular measure receiving a lot of attention is the concept of "Quality Adjusted Life Years." The idea behind Quality Adjusted Life Years is akin to asking yourself when faced with a fatal illness and an option of a medical intervention, "Would I rather live for two additional years in a hospital bed as compared to living normally for three months and then dying?" Sometimes surgery can have negative impact on quality of life. How can people pick between complex and risky options like these?
Furthermore, doctors are "rewarded" for good "outcomes," which sometimes has little to do with a patient's well-being. I have a friend whose father had heart surgery. The surgeon prescribed a very powerful medication to regulate his heart rhythm, but the drugs may be causing dementia. The drugs obviously had a very negative impact on the person's quality of life, but taking him off the drugs could have a negative impact on the surgeon's outcomes.
Tom Koch [1999, 2001] has published a couple of articles in OR/MS Today related to the dilemma of a national organ transplant service in the United States. The stated goals of such a system are "equality, efficiency and optimality," goals enshrined in law. But how do we determine what is equitable? For example, hearts and lungs will survive at most four to five hours outside a living host, so one could argue that proximity is optimal. But this policy creates inequities across the country due to imbalances in supply and demand, and in the location of major transplant centers. Therefore, equality, efficiency and optimality become conflicting objectives. There are more than 50,000 U.S. citizens waiting for a suitable donor. Who gets the organ? The person who has been waiting longest or the person who is the best match? Should younger recipients have priority since they will have more years of potential value? What is fair? Should the population of Alaska be guaranteed the same service as California? Keep in mind that these decisions are also highly influenced by politics.
How Can OR People Help?
In my experience, one of the major causes of inefficiency in the health care system is what I call "localized expertise." People working in the health care system are very knowledgeable about their own area but have relatively little understanding of what goes on in the next department. Doctors and nurses in the Emergency Department or in operating rooms do not really understand or sympathize with the problems faced by ward staff. People in hospitals have little appreciation for issues in long-term and home care. Occasionally, there are issues about "my work is more important than yours" or "my problems are bigger than yours." More often, it is simply too difficult for people to get a real handle on the whole "system." This is where OR professionals can play an important role.
For example, in one hospital where I consulted several years ago, the operating room management team had to decide how to allocate available time in the operating rooms over the course of the week. The team included nurses, managers and surgeons. They were very careful to be fair to each service and each surgeon. However, they did not consider the impact their decisions were having on the rest of the hospital. The allocation of surgeon time in operating rooms has dramatic effects on staff requirements in recovery rooms, wards, labs, medical imaging and administrative time. In particular, at this hospital, we discovered that the nursing workload in one of the wards was almost double on Wednesday compared to every other day of the week due to the weekly surgical schedule. By simply switching a few surgeons, we could effectively level the workload in all wards. Operating room planners did not have the tools to evaluate the impact that their decisions had on the rest of the hospital. In our case, we used simulation, but simpler tools could easily be designed.
The following examples of OR applications in health care are not intended as a comprehensive literature survey. I simply want to illustrate a few of the reasons why health care problems are unique and to demonstrate the wide variety of applications. For a more thorough review, readers should refer to the survey papers by Pierskalla and Brailer  or Jun, Jacobson and Swisher . Leonid Churilov and I have built a bibliographic database of about 800 papers describing OR applications in health care. This database is available in a searchable Web site at http://orchid.bsys.monash.edu.au/orchid/ and we welcome additions.
Simulation. Approximately 10 percent to15 percent of the papers in our database describe applications involving simulation. Obviously, one of the major issues in health care is waiting times (waiting for surgery, wait lists for transplants, location of emergency services, etc.), and most health care queueing problems are too complex to be analyzed theoretically. Therefore, simulation is a popular alternative. Simulation also helps people visualize the impact of local decisions on the whole system. One problem with using simulation in health care is the difficulty of collecting data. You cannot really follow patients around with a stopwatch. The health care environment also frequently involves multi-tasking; doctors and nurses look after several patients at once, and it is challenging to determine how to model their time.
Linear Programming and Goal Programming. LP has been used in a number of applications including staff scheduling, budget allocation and case mix management among others. Case mix is similar to the basic product mix example problem in every introductory LP text, except that it contains a few twists. The problem lies in deciding which set of procedures a hospital should perform to meet performance targets and stay within budget. One major obstacle in this process is that the hospital administration, unlike private industry, cannot dictate the case mix. As discussed earlier in the four worlds of health care, doctors are the gatekeepers. They decide what the hospital does, and they are generally more concerned about patient care than they are about the hospital's case mix issues.
Hospitals are usually divided into a "medical" and a "surgical" side. There is not much that we can do about the medical side; patients typically arrive with a variety of symptoms and must be treated promptly. However, the surgical side is primarily concerned with "elective" procedures. They are not elective in the sense that the patient has a choice about having them performed, but the patient and doctor will schedule them for some future date. These people are on the waiting list. Since we cannot dictate the "optimal" case mix, we simply determine whether or not a given hospital policy is feasible and point the hospital in the direction of feasibility.
Data Envelopment Analysis. There have been dozens of DEA papers published in the health care sector. For example, Kooreman  used DEA to compare the 320 nursing homes in the Netherlands. The homes were funded based on the number of beds and days of treatment. The author noted that it is hard to measure real health outcomes like "improved health status" or "improved quality of life," so he just used the output "number of patients treated" divided into four treatment groups.
The lack of reliable outcome measures is a common problem in health care. Most hospitals do not track the patient's health status after they leave the hospital unless they are re-admitted due to complications. Lack of data is even more pronounced in home care and nursing homes where data is typically manual and not standardized.
Integer Programming. There have been several papers published describing integer-programming applications, primarily for facility location and staff (nurse and physician) scheduling problems. A number of these describe locating emergency medical services, and ambulance location in particular. For example, Repede & Bernardo  developed a system for locating ambulances in Louisville, Ky.
According to U.S. standards, "95 percent of all (urban) ambulance calls should be served within 10 minutes." In their model, Repede and Bernardo assumed that the fleet size and the demand pattern changes over time. They provide a decision-support tool to help EMS planners relocate ambulances to maximize the total expected demand that can be served within 10 minutes. One of the distinguishing features of this type of facility location problem is that, once an ambulance is dispatched, it is no longer available to cover calls, and it could be out of service for an hour or more. Therefore, the fleet size is constantly shifting. This is a tricky stochastic problem that requires more attention.
Another application involves the optimization of radiation beams that travel through the body to treat cancer patients. These beams can travel through the body at a variety of directions and intensities. The objective is to maximize the radiation on the tumor and minimize the impact on healthy tissue, especially vital organs. Today, these calculations are typically done by hand. Linear and mixed integer models have been developed to improve patient treatment. (See for example Holder .) However, since these are considered medical interventions (they would actually be used on patients) the processes must be approved by the medical regulatory agencies.
AIDS Epidemic Modeling. Much of OR modeling in AIDS research is systems dynamics models. To quote from Kahn, Brandeau and Dunn-Mortimer  in their introduction to a recent special issue of Interfaces devoted to AIDS modeling: "The AIDS epidemic is a serious, growing public health problem worldwide, but resources for treating HIV-infected patients and for combating the spread of the virus are limited. Governments, public-health agencies and health-care providers must determine how best to allocate scarce resources for HIV treatment and prevention among different programs and populations. OR-based models have influencedand can influenceAIDS policy decisions. Mathematical modeling has had an effect on AIDS policy in a number of areas, including estimating HIV prevalence and incidence in the United States, understanding the pathophysiology of HIV, evaluating costs and benefits of HIV-screening programs, evaluating the effects of needle-exchange programs, and determining policies for HIV/AIDS care in California. Further work is needed to model a range of programs using comparable methods, to model overall epidemic control strategy, and to improve the usefulness of OR-based models for policymaking."
Queueing Models. There has been some work on managing hospital waiting lists and allocating beds in a hospital to various services. One interesting aspect of health care waiting lists, particularly for home care and long-term care, is the dynamic nature of the problem: as the queue increases, the reneging rate increases. People on the list either look elsewhere for service, become more seriously ill and go to a hospital, or perhaps die waiting.
Quality Management. In North America in the early to mid-1990s, hospitals were just beginning to do quality assurance using tools like statistical process control to monitor (immediate) outcomes. Manufacturing was at this stage in 1975, and other service industries were there in the 1980s. Other components of the health care industry (home care, nursing homes) are further behind. In contrast, the pharmaceutical industry was probably more in line with other manufacturing sectors.
One of the reasons for the delayed reaction in health care has been reluctance on the part of the medical community to acknowledge and report errors and problems. Physicians were often reluctant to even have their results tracked and benchmarked. There is a culture of silence in health care; they do not want to admit that mistakes can happen. The Harvard Medical Practice Study (Brennan et al ) reviewed more than 30,000 hospital records in New York State and found injuries from care itself ("adverse events") to occur in 3.7 percent of hospital admissions, more than half of which were preventable and 13.6 percent of which led to death. If these figures can be extrapolated to the 33.6 million admissions to American hospitals in 1997, then more than 98,000 Americans die each year as a result of preventable errors in their hospital care (Kohn et al ). By comparison, 97,860 people died in 1999 due to all unintentional accidents, which would make medical error the fifth highest cause of death (from the U.S. National Center for Health Statistics, www.cdc.gov/nchs). The analysis and prevention of adverse medical events has become a major focus of attention. In many situations, redesigning the processes can prevent errors.
What Can You Do?
Whenever I talk about OR in health care, a group of eager students will come up to me afterward very excited about trying to do something. They want to know how to start. My advice to them is that everyone knows someone who works in health care. That's how I got started. You simply chat with people in the business (doctors, nurses, managers, IS people, etc.) and tell them about what OR people can do and have done. Feel free to quote my examples. Do not expect them to immediately embrace your foreign concepts. However, I guarantee that you will quickly find someone who is frustrated with the inefficiency of health care and very willing to try anything, including quantitative methods. I would encourage you to start with something simple that you know will be successful. Given half a chance, we can make a big difference, but don't start with a complex mathematical model that has little chance of being implemented or that will take three years to implement. We need to demonstrate to people in the industry that we can make a difference. Professors should get their students involved in projects at hospitals. OR professionals should volunteer their services as consultants or on their local hospital board.
Everyday, when I get my morning newspaper, I read complaints that our health care system is not sustainable. I want you to think for a minute about what that means. Politicians complain that government cannot afford the annual increase in tax dollars that go to health care. Insurance companies complain that costs are out of control. Well, the bottom line is that you are going to pay for it, either in taxes or insurance premiums or out of your pocket. The problem is not going to go away. The culture in North America and elsewhere is that we do not act until we have a real crisis. I hope you realize that we are in a health care crisis. Baby-boomers are going to strain the system beyond its limits, and the escalating costs of drugs and technology are going to bring the system to its knees. I believe that OR can help, but it will take years to make the significant changes required to get through this crisis. What are you waiting for?
Michael Carter (email@example.com) is a professor at the University of Toronto, Canada. He is Director of the Health Care Productivity Research Laboratory in the Department of Mechanical and Industrial Engineering. He has supervised more than 50 student projects in hospitals, home care and mental health institutions. The author thanks John Blake of Dalhousie University for his valuable assistance with this article.
OR/MS Today copyright © 2002 by the Institute for Operations Research and the Management Sciences. All rights reserved.
Lionheart Publishing, Inc.
506 Roswell Rd., Suite 220, Marietta, GA 30060 USA
Phone: 770-431-0867 | Fax: 770-432-6969
Web Site © Copyright 2002 by Lionheart Publishing, Inc. All rights reserved.