An opportunity to make a difference

Operations research community well-positioned to impact healthcare around the world.

By Sheldon H. Jacobson

In order to have an impact on healthcare, operations researchers need to immerse themselves in the domain and dismiss preconceived ideas.

In order to have an impact on healthcare, operations researchers need to immerse themselves in the domain and dismiss preconceived ideas.

INFORMS Healthcare 2011 was a success. More than 500 people attended the conference held last summer in Montreal. The presentations crossed numerous areas of healthcare and medicine, including treatment optimization strategies, operations performance improvement and policy analysis. The analytics expertise that the operations research community can bring to healthcare has enormous potential to impact such systems around the globe. However, with such opportunities come certain responsibilities and a number of risks.

When provided with the opportunity to apply analytics to healthcare problems, it is easy to fall back on one’s own personal experiences with their physician or knowledge acquired through social contacts (such as a neighbor who happens to be a nurse). Although these interactions are useful, they rarely provide a complete picture of the challenges faced within the healthcare system, nor are they adequate to secure the necessary domain knowledge to maneuver through a healthcare problem. For operations researchers to fully appreciate the subtleties and nuances of healthcare – and learn the culture of healthcare systems – an effort to immerse oneself in this domain and a willingness to dismiss preconceived ideas (which are often flawed) is necessary. Although this will most certainly take most of operations researchers out of their comfort zones, the end results will lead to opportunities to make a difference. As the Turing test can be used to determine whether a qualified judge cannot distinguish between a human expert and a computer, a Turing test for those working in healthcare analytics is whether a physician or healthcare professional cannot distinguish between a physician or an operations researcher when discussing domain knowledge in the specific healthcare field.

Numerous journals publish research within the operations research community. INFORMS publishes several that hold very high reputations. Researchers read articles that are relevant to their own areas of interest or areas that they wish to learn more about. It is critical to keep apprised of the latest technical developments or new areas of application where operations research methodologies have been successfully applied. In contrast, physicians caring for patients read medical journals to learn better clinical practices so that they can provide the most effective treatment plans and care for their patients. This alternate focus brings to light why physicians recertify their clinical skills every decade and show that they are clinically current and competent by the most up-to-date standards of care.

Operations researchers often focus on a system and its process, searching for a better if not the best solution to a well-defined model. As a result, the impact of analytics to date has been focused around those aspects of healthcare that match well with O.R. models and tools. For example, O.R. methodologies are well-suited to suggest system changes that could lead to a 20 percent improvement in hospital resource utilization or a 25 percent reduction in costs based on a redesigned staff schedule. This is something that operations researchers do well. However, not all healthcare issues lend themselves well to O.R. methodologies, given the issues of concern and the decision-making process. To illustrate this point, physicians are most concerned with patient outcomes, which is why robustness rather than optimality may be a better objective in healthcare. This explains the recent decision by the NIH National Heart, Lung and Blood Institute to prematurely terminate a study on the use of high-dose, extended release niacin in conjunction with a statin to reduce cardiovascular events. In this study, niacin produced the desired biochemical markers (i.e., higher level of high density lipids, or HDL), but this did not translate into a positive clinical outcome (i.e., lower cardiovascular events).

Any system that deals with human lives and decision-making will be fraught with uncertainty. There may be hidden reasons why bed or staff utilization levels gravitated to suboptimal levels, which are not apparent to even the most skilled operations researchers. This is also why healthcare research conducted by operations researchers demands that medical expertise be involved throughout the entire study process. What may be obvious to a physician may be highly counterintuitive to an operations researcher. Contact with health services and information technology personnel will also be critical, given the growth of electronic medical records and their ubiquitous spread throughout the health delivery system. Therefore, systems improvements that move toward both optimality and robustness are more likely to be well-received by the healthcare system.

When new medical devices, procedures or treatments are created, they must go through a rigorous series of clinical trials, the results of which are evaluated by the Food and Drug Administration (FDA). These trials occur in three phases: 1. focusing on safety, 2. focusing on efficacy, and 3. focusing on benefit. The purpose of such a rigorous evaluation is to assert that the device, procedure or treatment will produce the clinically beneficial effect across a broad spectrum of the targeted population. As operations researchers move into the healthcare domain, they will need to understand their role in this process. In essence, analytics may serve as a Phase 0 in such studies, identifying the potential or the opportunity for new approaches to address existing problems. However, where modeling and analysis ends is where those in the medical community can begin to consider how clinical trials must be designed to move analytics from the computer into the clinic. To achieve such a transformation will require operations researchers to become more knowledgeable about the entire approval process, even though the final result may be far removed from where their major contributions may end.

A small group of medical students opts to pursue a more research-focused medical track by doing both an M.D. and a Ph.D. These medical scholars must satisfy the requirements of both degrees, often spending as many as eight years in school before they even consider their medical residency. The fields of their Ph.D. degree can be quite varied, though many pursue degrees in biology, biochemistry, biophysics, biomechanics and material sciences, which naturally lend themselves to medical bench research. Most medical students are unaware of the O.R. field, and hence, do not consider a Ph.D. in operations research. This may also be tied to their limited background in mathematical modeling and analytics. An untapped opportunity may exist for operations research Ph.D. programs to expand their reach into the medical community through M.D./Ph.D. programs. Gradates of such a program could serve to bridge O.R. work to the broader medical community more effectively than those holding just a Ph.D. degree.

The potential for operations researchers to undertake successful translations of O.R. research into the healthcare industry is enormous.

The potential for operations researchers to undertake successful translations of O.R. research into the healthcare industry is enormous.

Dissemination of healthcare research within the operations research community may inform O.R. colleagues of an individual’s research activities, but it does little to bridge and translate O.R. work into the medical community. Medical journals are not interested in the technical details of O.R. models and algorithms. Instead, they are interested in new insights that can be gleaned from such analytics that can lead to improved medical practice and/or better-informed healthcare policy and decision-making. Preparing articles for medical journals requires a rethinking of the focus of O.R. work and the target audience. By writing manuscripts in this manner, the final document may appear rather limited on methodological content. The key is to succinctly describe the problem addressed, the approach taken, the discussion of the results, the limitations of the study and salient conclusions to be drawn. One of the most read medical journals is the New England Journal of Medicine (NEJM). In Web of Science 2010, the NEJM has two-year and five-year impact factors of 53 and 52, respectively. In contrast, Operations Research has two-year and five year impact factors of 2.0 and 2.7, respectively. Moreover, medical writers and journalists review the NEJM to report on the journal paper’s findings. As operations researchers make further advances into the healthcare domain, the opportunity to disseminate O.R. work directly into the medical community by publishing its very best findings and insights in medical journals is critical to the success of the O.R. community’s efforts.

As noted, INFORMS Healthcare 2011 was a success. Of the 500+ registered attended, more than one-third were students, which paints a bright future for the application area within the O.R. community. On the other hand, very few physicians attended. The O.R. community’s future success requires operations researchers to move their research efforts into the medical community and participate in medical conferences. The INFORMS and Society for Medical Decision Making conference to be held in Phoenix in November 2012 is a welcome step in this direction. The annual Mayo Clinic Conference on Systems Engineering & Operations Research in Health Care and Modeling for Public Health Action – From Epidemiology to Operations, sponsored by the Centers for Disease Control and Prevention (CDC), are excellent examples of medical organizations embracing operations research, modeling and analytics within their organizations. Operations researchers have been key drivers in initiating these events (Brian Denton, who was at the Mayo Clinic prior to joining North Carolina State University, with support from Mark Hayward and John Osborn, and more recently, Tom Rohleder, for the Mayo conference; and Michael Washington for the CDC conference).

Given the broad interest in healthcare, many major medical conferences also have teams of medical writers assigned to cover the conferences and report on the new medical findings discussed at the events. In contrast, it is challenging to get journalists to attend O.R. conferences and report on what operations researchers are doing. The success of the O.R. community’s efforts in healthcare will depend on its attendance and participation in medical conferences so that the results of O.R.’s research can be injected into the very community that it is positioned to help and to translate that research into clinical practice and better healthcare outcomes.

The potential is enormous for operations researchers to undertake successful translations of O.R.’s research into the healthcare domain. The CDC has indicated a strong interest in operations research and how it can inform public health policy, as evidenced by the growing number of operations researchers on its staff. However, with such great rewards come risks. Although healthcare costs are becoming a growing concern, patient outcomes will always play a primary role in the design and delivery of healthcare systems. If O.R. research produces poor patient outcomes, resulting in lives lost or impaired, the credibility of O.R. efforts across the entire spectrum of healthcare will be damaged. Although O.R. analyses often focus on efficiency, cost and utilization metrics, medical objectives such as quality adjusted life years (QALY), which has been the standard to measure cost and benefit within the healthcare domain for many years, must be considered. Operations researchers must take this necessary step to incorporate health-oriented measures into their efforts to gain further credibility and ease the transition of their findings into the healthcare community. Also, given that the National Science Foundation program in Service Enterprise Engineering funds a significant amount of healthcare-related research within the operations research community, the need to involve physicians and medical personnel in such projects is critical if such investments will pay the desired dividends. Moreover, given the data needs of such efforts, the ability to secure such information from the medical community and share it across the operations research community is imperative for the success of such projects.

The operations research community is positioned to impact healthcare around the world. The key asset that the O.R. community brings to the healthcare field is the manner in which operations researchers approach and analyze problems. The O.R. community’s analytical thinking provides the greatest opportunity for its efforts to meet the challenges faced by the healthcare system. The O.R. community’s effectiveness in this role will be a function of its willingness to step out of traditional comfort zones for publication, dissemination and participation and boldly immerse itself in this domain of study. Such steps drive the success of the efforts and impact that O.R. will make in shaping a better healthcare system for all.

Sheldon H. Jacobson (shj@illinois.edu) is a professor in the Department of Computer Science at the University of Illinois at Urbana-Champaign. He has researched a variety of issues in the public health domain since 1996, including pediatric vaccine pricing and stockpiling economics, the relationship between transportation and obesity, and the impact of cell phone laws on automobile safety. His findings have been disseminated both within the engineering (e.g., Operations Research) and medical (Journal of the American Medical Informatics Association) literature, and he has received funding from the National Science Foundation for this work.

Acknowledgements

The author wishes to thank the many people he interacted with at INFORMS Healthcare 2011 for their spirited discussions and interactions. Special thanks to James Benneyan (Northeastern University) and Russell Barton (Penn State University and NSF) for their comments on an earlier draft of this article. The author would also like to thank Janet A. Jokela, M.D., M.P.H. for her feedback on this article and for being his wonderful wife.