When corporate executives, health care leaders, and policy makers discuss the challenge of curbing U.S. health care costs, the conversation invariably turns to the sickest 5% of the population, who consume 50% of health care spending.
For a long time the hope has been that improving the efficiency and quality of their treatment would significantly reduce the $3.5 trillion that the United States lays out annually for health care. Over the past two decades this thinking has led employers, insurers, and health systems to embrace expensive disease-management programs that, operating in parallel with patients’ primary-care physicians, use registered nurses and social workers to monitor, coach, and provide services to many people in the top 5%. While these programs do increase the quality of their care, our health system, Kaiser Permanente (KP), and nearly all others have found that they do not reduce net costs.
To learn why, KP started looking at internal studies of utilization and detailed information on care given its 4 million patients in Northern California, which had been captured by its electronic health record (EHR) system. KP’s clinical researchers made two discoveries. First, the makeup of the most expensive 5% is much more heterogeneous than has been appreciated. The group comprises three roughly equal segments of patients with very different medical needs: people with one or more chronic medical conditions that could be improved or kept under control; people who suffer a onetime catastrophic health problem; and people with severe chronic conditions who can’t be returned to good health and require expensive treatment continually. Second, the people in the first two segments of the top 5% change unpredictably from year to year. All this explains a lot about why disease-management programs haven’t delivered positive returns: They aren’t designed to address the heterogeneity and unpredictability.
In parallel, KP developed a new model for treating people with multiple but relatively manageable chronic diseases—focusing on both those who are currently in the top 5% and those who could end up there in the coming years if their medical problems worsen and their health deteriorates. We believe that addressing this entire group of patients presents the biggest opportunity for improving outcomes and increasing savings. Our approach uses technology and relatively inexpensive medical staff to provide expanded support to primary-care doctors so that they can oversee and address the chronic needs of patients directly instead of relying on largely independent disease-management programs.
To date, we’ve implemented it in California, Virginia, Maryland, and the District of Columbia, and we’ve found that it significantly reduces costs and improves the quality of care. Despite the investments required, the leveraged-primary-care model has increased KP’s operating margins, enabling it to offer nearly 5 million members of its network premiums that are 10% to 15% lower than its competitors’. The higher margins have also allowed KP to fund major capital investments (in excess of $1 billion a year), to fulfill its obligations to care for the under- and uninsured, and to finance training for the next generation of doctors. KP’s plans in Northern California and the mid-Atlantic region now consistently rank among the top five health plans in the nation for quality, and we believe the new model is one of the reasons why.
There is strong evidence that the model can be broadly adopted. Recognizing that KP’s HMO structure, which combines an insurance plan and a health system, is unusual in the United States, one of us (Robert Pearl) decided to study three other primary-care organizations that use a model like KP’s but don’t have their own insurance plans. They operate in 17 states and collectively serve more than 250,000 people. These organizations, he found, had also achieved excellent quality at lower costs. Patients experienced fewer complications from chronic illnesses and avoided developing new ones and, as a result, visited emergency rooms and hospitals 20% to 50% less often. We believe this demonstrates that a leveraged-primary-care model could be used by any organization whose reimbursement is dependent on outcomes of care—rather than simply the volume of care provided.
In this article we’ll share what we learned about the segments that make up the most expensive 5% of the patient population. And we’ll describe our model in more detail, in the hope of encouraging other medical groups to adopt a similar approach and employers to demand it.
The Three Cohorts
When health care experts talk about the most expensive 5% of patients, they usually have in mind people suffering from chronic conditions such as diabetes, stable heart failure, asthma, and mental illnesses. These patients go to emergency rooms often, see doctors frequently, and require periodic hospitalization. The aim of disease-management programs is to improve their health—and reduce medical costs—through better care and coaching. But what the experts don’t realize is that these people account for only a third of the most expensive 5% of patients and that the makeup of this cohort isn’t constant; it changes from year to year.
Another third of the top 5% are patients who experience a onetime catastrophic health event: a major trauma, an extremely premature birth, or a sudden life-threatening illness such as acute cancer. The care for this second cohort consumes about 35% of the total spent on the 5% in a single year. But since most onetime events can’t be predicted, there’s little opportunity to achieve savings. Moreover, many of the people who are in this cohort one year will drop out of it the next as they recover or die.
The final third of the 5% are people with medical conditions such as severe heart failure and chronic renal disease, who require expensive, ongoing treatment every year. Many were in the first cohort years before but deteriorated, and with few exceptions (such as people who receive kidney, lung, and heart transplants), little can be done to restore them to good health or avoid large future medical bills. The people in this cohort also account for about 35% of the total dollars spent on the 5%. Although disease-management programs can make their care more efficient and effective, most efforts to improve it have had a minimal net financial impact, because these patients need especially intense interventions and costly medications and therapies. In addition, most of them will require extensive and extremely expensive end-of-life care in the near future.
The bottom line here is that the actual opportunity to reduce the costs of caring for the most expensive 5% of patients is much smaller than is appreciated. The only cohort whose costs can be significantly reduced is the first, but these people account for just 30% of all the money spent on the top 5%. And the fact that the people in this cohort change every year means that to be truly effective, cost-savings programs need to target patients who might end up in it too. That’s a huge group. Consider the nation’s 55 million Medicare members over the age of 65. Roughly 27.5 million have five or more chronic conditions that could be kept under control with proper care, but only 4 million or 5 million of them end up in the first cohort of the top 5% in any given year.
Because the rest are also at high risk of landing in this cohort the following year, a care program would have to include all 27.5 million to achieve maximum success. For conventional disease-management programs, the cost of this would be wildly prohibitive. KP’s leveraged-primary-care model obviates that difficulty by integrating the support staff directly into the primary-care doctor’s practice, using people who command lower pay than registered nurses and social workers do, and by employing technology. Here are the key differences between the two approaches.
Leveraged Primary Care vs. Disease Management
Disease-management companies typically organize nurses and social workers into teams that focus on just one disease, which means that patients with several chronic illnesses could be dealing with multiple teams. These interactions occur outside the primary-care physician’s practice and often duplicate its work: Both the teams and the physicians monitor each patient’s laboratory results, medications, and overall health. Most patients like the convenience of disease-management programs, which they can call on for help navigating the complex care-delivery system, scheduling visits with doctors, arranging transportation, and connecting with social services. Many also take advantage of the educational opportunities and coaching that such programs provide. But the dollars the programs consume could be better used to address the needs of far more individuals.
This expensive approach does make sense for patients with complex and persistently high-cost conditions such as end-stage kidney disease and severe cardiac or pulmonary problems, who are at constant risk of deteriorating and needing hospitalization. Avoiding just one ICU admission each year can cover a disease-management program’s expense. Like other organizations, KP uses such programs to increase the quality of care for those patients.
But, again, applying this model to all people with multiple chronic diseases, or comorbidities, simply isn’t economically feasible. An alternative model that many have tried—using registered nurses or nurse practitioners in the doctor’s office to help care for patients with chronic diseases—also doesn’t work. It, too, involves employing relatively highly paid people to perform many tasks that don’t require their skills: screening databases, arranging laboratory services and radiology tests, obtaining information on each individual’s subjective view of his or her health, asking whether the doctor’s plan is being followed, and extracting the data needed to see if outcomes match physicians’ expectations.
KP’s model solves the problem of reaching a large population cost-effectively, by incorporating four components:
Integration with ongoing primary care.
This avoids the duplication of services. But doctors still get support with outreach, which allows them to leverage their expertise in the most efficient way possible.
The model uses four types. One is a comprehensive EHR system that provides physicians up-to-date information on each patient’s condition, highlighting the need for additional treatments or testing. The system makes it easy for the physician and support staff to quickly order tests, change medications, send messages to patients, and monitor clinical results. It can also be used to find and prioritize individuals with chronic diseases who need attention before a routine visit. A second type is wearable devices that record data on blood pressure and weight. A third is computer-generated voice and text messages that remind patients about preventive screenings. And a fourth is a smartphone video function that allows doctors to monitor patients after they leave the hospital, immediately address new medical problems, and avoid readmissions.
Low-cost medical assistants.
At KP each primary-care doctor is assigned a medical assistant. By communicating with the patient, reviewing information, readying it for the physician, relaying physician orders, and arranging transportation through community services, medical assistants help physicians manage patients’ chronic diseases better—and at a small fraction of the cost of other approaches. They save doctors enormous amounts of time—just extracting all the data from electronic health records, assembling it, and organizing patients’ clinical information could take several hours a day. While their pay varies by geography and union agreements, assistants make $40,000 to $50,000 a year, roughly half what registered nurses or nurse practitioners earn. Their training takes a year and is offered by community colleges, state colleges, and vocational schools.
Added recently to the model, pharmacists can access laboratory data and, using protocols created by the physicians, make many of the medication changes needed, saving doctors even more time.
The new model enables each physician to economically provide continual care for about 1,800 patients—30% of whom have one or more chronic diseases—that’s of much higher quality than conventional primary care. The doctor knows exactly what the medical assistant is doing, the specific drugs the pharmacist is titrating (fine-tuning the dosage of), and the preventive services the IT voice and text systems will encourage. The model also allows doctors to tailor visits to patients’ needs: Rather than having everyone with chronic conditions automatically come in for a checkup every three to four months, they might schedule some individuals to return monthly for a longer visit and others whose lab results are in line with expectations to come in only once a year.
Although KP’s program depends heavily on its sophisticated EHR system and a variety of internally developed computer applications, these types of analytic tools are increasingly available through commercial software vendors. A minimal investment in medical assistants and analytic software can generate savings (in the form of fewer emergency room visits and days in the hospital) that are a multiple of the dollars spent. Once implemented in a capitated health care organization—in which providers receive a fixed payment that covers each patient’s expected health care services over a defined period and are also held accountable for high-quality outcomes—the model creates a virtuous cycle. Relatively low investments in primary-care practices improve the health of large populations of patients and generate progressively greater savings each year, which can be used to fund additional medical assistants and technology the next year.
The leveraged-primary-care model isn’t only cost-effective. Consider some of its successes in addressing these life-threatening, expensive diseases:
Fifty percent of patients who die today from colon cancer—usually after running up massive medical bills (often more than $200,000)—could have survived or even avoided the malignancy in the first place with proper screening. The majority of colon cancers take 10 years to progress from a benign polyp to an invasive cancer. If the polyp is identified and removed early, the malignancy never develops.
For the majority of people, precancerous and malignant polyps can be found through a simple stool test. When it’s performed annually, its detection rates are the same as a colonoscopy’s. The stool sample is obtained in the privacy of the individual’s bathroom without bowel prep or the risk of intestinal perforation. The challenge is getting patients to participate. The leveraged-primary-care model greatly increases the percentage of patients who do: KP’s screening rates are 90%—substantially higher than the U.S. rate of about 70%. As a result, while mortality from this cancer is declining nationally, for KP’s members it’s falling much more rapidly. When they visit their primary-care physicians, a medical assistant can check the EHR to see whether they need the test and, if they do, hand them a diagnostic kit. If a patient is in the appropriate age group but either didn’t come for care that year or failed to return the kit, a computer application identifies the gap and sends a voice or text reminder. If necessary, the medical assistant can request that a kit be sent to that patient.
High blood pressure.
According to an editorial by Naomi D.L. Fisher and Gregory Curfman in the November 6, 2018, issue of JAMA, high blood pressure, or hypertension, is “the single largest contributor to cardiovascular disease, causing stroke, heart failure, and coronary artery disease, and also is a major contributor to kidney disease.” The reasons it’s poorly controlled are complex but relate to two factors. First, it can take doctors a lot of time to fine-tune the correct dosage of hypertension medications for each patient. Second, many patients don’t take the prescribed medications consistently. A medical assistant can help physicians address the first issue by obtaining more-frequent blood-pressure readings in person or by analyzing data from remote monitors, and the second by calling patients or sending them email or text reminders. That’s the approach we take at KP, and as a result, 90% of our patients with high blood pressure have it under control, compared with only 54% across the United States.
The incidence of this chronic illness is growing, and that’s troubling, given that its complications can be grave and costly. When early foot and ankle irritations, abrasions, and ulcerations aren’t tended to, patients often end up requiring foot and lower-limb amputations. A medical assistant, inspecting potential problem areas via a smartphone’s video function, can encourage diabetes patients to be vigilant and immediately bring concerns to the physician’s attention, often weeks before the problem would otherwise have been identified during a routine visit.
More broadly, the KP approach—combined with other efforts to improve operational efficiency, maximize prevention and patient safety, and utilize “centers of excellence”—has allowed us to reduce hospitalizations for Medicare patients in California (where we own our own hospitals) to an average of 0.7 days per patient per year, which is half the national average. Before the model’s implementation, our rate was closer to 1.2 days per Medicare enrollee per year, or 85% of the national average.
Improvements in prevention have also reduced KP’s mortality rates. For example, KP members in California and the mid-Atlantic region are now 28% less likely to die from colon cancer, 40% less likely to die from sepsis, 43% less likely to die from heart disease, and 14% less likely to die from stroke than people in the U.S. population as a whole are. Previously, KP’s outcomes in these areas were only slightly better than the national averages and had not been near the best on either coast.
The Model Is Broadly Applicable
Many people assume that only large integrated organizations with multispecialty medical groups can use a model like KP’s. But as we noted, three smaller organizations have adopted one and seen patient outcomes and financial results similar to those we’ve had. All three are predominantly primary-care organizations with dozens of locations. All are paid on a capitated basis and focus on Medicare Advantage populations—especially the sickest enrollees, for whom Medicare pays a higher rate of reimbursement. They also serve many people who are dual eligible, or enrolled in both Medicare and Medicaid. At each organization, financial success depends on preventing complications and providing alternatives to hospitalization and trips to the emergency room—not on maximizing the volume of visits, procedures, or hospital stays.
Because their patients are much sicker than the typical Medicare enrollee, all three organizations assign fewer patients to each physician than the community average for primary care. They support primary-care doctors with office staff equivalent to KP’s medical assistants, promote more-frequent contact with enrollees, and use EHR systems to keep tabs on patients’ health, provide point-of-care decision support, and issue alerts and reminders. Also like KP, they use RNs and higher-paid case managers operating outside primary-care offices to coordinate the care of their hospitalized and sickest patients (those in the third cohort).
The savings will grow as outcomes get better for the entire population served.
One of the three, ChenMed, has 59 primary-care clinics in eight states. Because the needs of the people it serves are so complex, it makes building a strong relationship between doctor and patient a high priority. In 2018 its physicians spent, on average, 189 minutes with each patient, compared with an average of only 20.9 minutes for all U.S. primary-care doctors. Even when those averages are adjusted to account for the increased risk of ChenMed’s population, its physicians devoted about twice as much time to seeing their Medicare patients as their counterparts in typical community-based office practices did. And that extra time appears to be well worth it: In comparison with national averages for other providers, ChenMed has 34% fewer emergency-room visits and 26% fewer hospital days per enrolled patient per year, and 16% fewer hospital readmissions per 1,000 Medicare members. Medicare Advantage has given ChenMed’s plans ratings of four to five stars (out of five) for quality and patient satisfaction. (Full disclosure: ChenMed paid one of us, Robert Pearl, for a day of consulting two years ago.)
ConcertoHealth, which operates in Ohio and Washington, has an especially strong focus on individuals who, because of their severe medical problems, low income, and disabilities, are eligible for both Medicare and Medicaid. More than 80% of ConcertoHealth’s patients have at least three chronic diseases, and nearly 50% have a mental health diagnosis. Its leveraged-primary-care model has pushed its hospital utilization and readmission rates and emergency room visits below the national averages for Medicare patients, by 16%, 50%, and 17%, respectively. Medicare Advantage has given its plans four-star ratings.
Iora Health, which has 49 primary-care practices in 10 states, has seen similar reductions. According to 2018 data, patients who had been in its leveraged-primary-care program experienced a 15% reduction in hospital utilization after one year, a 24% reduction after two years, and a 42% reduction after three years. Iora’s hospital utilization and readmission and emergency department usage rates are 40%, 36%, and 23% below the national averages for Medicare patients, and Medicare Advantage has given its plans 4.4 stars.
We believe that in the years ahead KP’s strategy of supporting primary-care doctors with medical assistants who are armed with advanced IT will save even more money. It can be applied to patients with any chronic health-threatening problem. Take cancer and type 2 diabetes. Smoking and a poor diet and lack of exercise are commonly identified as their leading causes, but in typical primary-care practices, doctors rarely have time to discuss those habits with patients in depth, steer them toward a more healthful path, and help them stay on it. But in a leveraged model, added support staff and information technology make those interventions possible. Also consider the large gap between optimal results and current patient outcomes in areas such as hypertension, blood lipid control, asthma management, and diabetes. Broad adoption of the KP approach, which data from the National Committee for Quality Assurance indicates is producing significantly better results, could help close it. And while the greatest value of the approach is its ability to improve the care of patients with multiple chronic diseases, it also can help increase the effectiveness of disease-prevention efforts for healthy individuals, extend lives, and reduce future costs.
A major obstacle to progress, however, is the fee-for-service payment system that still dominates U.S. health care and fails to adequately reward efforts to prevent diseases and complications. Invariably, fee-for-service reimbursement leads to a higher volume of interventions—but not improved health outcomes. To increase the quality, convenience, and affordability of medical care, the country must rapidly move toward a capitated reimbursement system. As KP and other health systems have proved, capitation spurs the adoption of innovations that improve outcomes, reduce costs, and generate the funds needed to develop more innovations. (See “The Case for Capitation,” HBR, July–August 2016.)
Medicare has taken a number of initial steps to move from paying providers solely for the volume of procedures they perform to providing incentives and penalties based on the outcomes they achieve (which is known as value-based payment). This includes new programs for primary-care physicians announced in April 2019. If this effort gains momentum and large employers pressure private insurers to follow suit, it will make investments in a leveraged model much more attractive for primary-care physicians. Until then, the funding will have to come from the payers: insurance companies, employers, and state and federal governments.
Out of desperation to reduce their health care costs, some large employers and many health plans have contracted with disease-management programs, only to be largely disappointed with the results. It’s time for large employers and insurance companies to change course. By funding more medical assistants and paying for IT systems similar to KP’s, purchasers and insurers could see bottom-line improvements within a few years. The savings will grow even greater over time as outcomes get better for the entire population served and the total cost of caring for the sickest 5% of patients falls.
With the U.S. population aging, the incidence of chronic disease will continue to rise, driving up health care costs dramatically. The operational and capitated-payment approach used by KP and the three other medical groups offers a way to address this growing crisis.
A version of this article appeared in the January–February 2020 issue of Harvard Business Review.
Robert Pearl, MD, is the former CEO of the Permanente Medical Group, a physician organization, which, along with Kaiser Foundation Health Plan and Hospitals, is part of Kaiser Permanente. He is also a professor at Stanford University’s medical school and a lecturer at its business school. He is the author of Mistreated: Why We Think We’re Getting Good Health Care and Why We’re Usually Wrong.
Philip Madvig, MD, is the former associate executive director of quality and hospital operations at the Permanente Medical Group.