One major opportunity for lowering health care costs without compromising outcomes is to persuade physicians and patients to use sites that provide high-quality services such as lab, tests, MRI scans, and procedures at a lower cost than others. But one challenge is changing ingrained practices or habits. Only by taking that challenge into account can payers and health systems estimate the potential savings, which affects the design and investments in such initiatives. This article describes a model developed by the Healthcare Transformation Institute at the University of Pennsylvania and Embedded Health Care to realistically quantify the potential cost savings by changing practice patterns.
In the constant battle against rising health care costs, payers and health systems are recognizing the importance of the site of service — for example, whether your MRI scan is performed in a hospital, a doctor’s office, or a freestanding facility. Indeed, the place where care is delivered heavily influences both health care prices and variation in prices.
The appeal of this lever as a cost-savings mechanism is obvious: Adjusting the site of service can yield savings, usually without sacrificing quality of care. Many of the analyses in this area have predicted that massive savings potential exists. However, efforts to shift sites of service have not realized such savings. One potential reason is that these estimates assume all high-cost behaviors are equally easy to change. They fail to consider the challenges of getting physicians to alter their current referral practices and either send their clinical work to lower-cost sites or recommend that patients use them.
The Healthcare Transformation Institute (HTI) at the University of Pennsylvania and Embedded Healthcare (EHC) developed the Clinician Behavior Change Model to realistically quantify the potential cost savings by changing practice patterns. In this article, we outline our approach to model potential savings from optimizing site of service for laboratory tests. This approach is already applicable and is being used by HTI and EHC to estimate site of service pilots for a broad set of clinical activities, including radiology, surgical procedures, and specialist referrals. These realistic savings estimates can better inform the business case for investments in any number of new payer initiatives designed to influence the site of service, including new payment models, incentive programs, or electronic-medical-records-based tools (e.g., decision-support tools or “best practice advisories”) deployed at the point of care.
In general, the model weighs the potential savings from a successful switch against a set of counterbalancing factors that address how difficult it is to change clinician behavior. Barriers such as contracts, strong patient preferences, clinical expertise, and the size of a financial incentive paid to the physician for optimal referrals all factor into whether we can turn a high-cost lab-panel draw into a lower-cost one. When linked together, the model illuminates how much actual opportunity there is to reduce costs by changing site of service. This model is currently being used by a large national insurer, HTI, and EHC to design new clinician-facing incentive programs geared toward motivating and rewarding clinicians who successfully shift to higher-value sites of service. The insurer intends to launch at least four pilot efforts in 2021 that are aimed at tapping opportunities identified with the assistance of the model.
The model. To start, we look at gross clinical volume and cost, which establishes the baseline of utilization. We then consider the proportion of that activity for which there is a lower-cost, clinically equivalent alternative available (“total low-value volume”).
We then determine the volume of tests that could be influenced with an intervention by filtering out those that can’t be changed, yielding the intervenable volume. For example, we subtracted lab tests that required specialized, expert facilities (such as ordering Interleukin 6, or IL-6, levels) because of their clinical complexity and the relatively few alternative lab sites that could perform them. We also took into account whether a practice is independent or owned by a health system and whether it has previously used an alternate lower-cost option. Clinicians whose practices are owned by a health system may have less freedom to choose a laboratory; the health system may mandate its own facilities and clinicians be used.
Even if clinicians can change their low-value practices and can be motivated to do so (by being offered a financial incentive), it would be impossible to move all intervenable volume to lower-cost sites in a short time due to ingrained practice patterns stemming from clinicians’ habits and patients’ preferences. The model estimates what percentage can realistically be moved — the targeted volume of clinical activity.
Getting this target right is important. If the program sets a goal that physicians feel is unattainable in the given timeframe, they may disengage from the objective or refuse to participate even before the program is initiated. Thus, it may be useful to set an initial, easily attainable target and a series of increasingly more ambitious targets, paired with a set of corresponding incentives. For example, if a clinician is currently sending 20% of laboratory orders to a low-cost facility, it may be unreasonable to expect him or her to increase that proportion to 70% in year one, given the existing contractual relationships or entrenched patient or health system preferences. It would be more realistic (and motivating) to set an initial lower goal of, say, 30% or 40%, for the first year and increase that to 50% to 60% in the second year, and 70% to 80% in the third year.
The targeted volume also accounts for clinicians who choose not to participate. Multiplying the targeted volume by the unit savings between the high- and low-cost clinical activity and then deducting the costs associated with the program — such as financial incentives paid or the revenue that payers lose from non-financial incentives (e.g., waiving certain prior authorization obligations for drugs or devices) — generates the realistic net savings from a program aimed at persuading physicians to use or refer patients to lower-cost sites of service.
A case study. Working with a large national insurer, we applied the model to clinicians’ ordering of outpatient laboratory tests in one market for one calendar year.
The clinicians were sending 174,788 orders to low-value labs, which cost a total of $22.6 million. After excluding complex lab tests, those performed by providers affiliated with health systems, those taking place at an in-office lab, and so on, only 42%, or 73,432, of the orders remained. They are the intervenable volume.
We designed the model to target 50% of all intervenable lab tests in Year 1. Analysis of the insurer’s data on patterns for participating physician practices showed that, on average, physicians regularly referred patients to at least three distinct laboratory facilities. The model assumed physicians might be able to move 50% of the high-cost lab volume once patient preferences or the proximity of the lab sites to patients’ homes were factored in.
In addition, we estimated that assigning 15% of the total estimated savings to physician incentives would result in the movement of 40% of eligible lab orders to a higher-value, lower-cost location. Intervenable volume x 50% targeted x 40% ultimately converted is the targeted volume. After running the model with these adjustments, we estimated that approximately 15,000 (8.4%) of lab orders could be changed to lower-cost facilities in Year One if we spent $259,995 on incentives, which would generate a realistic net savings of $1,473,000 (6.5% of the total that was being spent). That meant there would be a return on investment of $9.65 for every dollar spent on incentives to clinicians. These projections will be tested in 2021, when the insurer carries out the pilot projects.
The lessons. There are three important lessons from applying this model.
Actual savings from changing site of service, or any physician behavior, will be significantly lower than the total low-value volume or the intervenable volume suggests. By identifying clinicians who are more and less amenable to changing their referrals to certain sites of service, the model allows payers to focus efforts on those most likely to change, which may produce a higher return on investment in the behavior change program.
The model can be used to shape a payer’s short- and long-term strategies for reducing costs. By this, we mean the mix of easier-to change and harder-to change behaviors that can be targeted over time. For example, getting physicians to move their lab orders from a high-cost site to a low-cost site may represent a low-hanging fruit that generates savings in the short run. But even though the potential savings from persuading physicians to refer patients to lower-cost specialists or sites where surgical procedures are performed might be greater, achieving that is harder and may take longer due to the more ingrained nature of surgical referrals, the clinical complexity of these actions and outcomes, the need for specialized operating rooms, adjustments in physician workflows, and physician admitting privileges at hospitals.
The estimates from this model will become more accurate as its use increases because data can be refined by using data from the actual experiences of implemented programs in specific areas.
In striving to simultaneously improve outcomes and reduce costs, it’s crucial for players in the health care system to have a deep understanding of what’s achievable, how to overcome the barriers, and how long it will take to attain the goal. Our model can help provide that knowledge.
Ravi B. Parikh, MD, is an assistant professor in the Department of Medical Ethics and Health Policy and Medicine and a senior fellow at the Healthcare Transformation Institute at the University of Pennsylvania.
Connor W. Boyle is an analyst at the Healthcare Transformation Institute and a research coordinator at the University of Pennsylvania.
Emily Roesing is a senior director of client services at Embedded Healthcare.
Justin E. Bekelman, MD, is a professor at the Perelman School of Medicine and director of the Penn Center for Cancer Care Innovation at the University of Pennsylvania.
Amol S. Navathe, MD, is the co-director of the Healthcare Transformation Institute and associate director of the Center for Health Incentives and Behavioral Economics at the University of Pennsylvania. He is also a staff physician at the Philadelphia VA Medical Center.
Ezekiel J. Emanuel, MD, is the vice provost for global initiatives and co-director of the Healthcare Transformation Institute at the University of Pennsylvania.
Copyright 2020 Harvard Business School Publishing Corporation. Distributed by The New York Times Syndicate.