Time to tackle unwarranted variations in practiceBMJ 2011; 342 doi: https://doi.org/10.1136/bmj.d1513 (Published 17 March 2011) Cite this as: BMJ 2011;342:d1513
- John E Wennberg, Peggy Y Thomson professor emeritus in the evaluative clinical sciences
- 1Dartmouth Institute of Health Policy and Clinical Practice, 35 Centerra Parkway, Lebanon, NH 03766, USA
Since Alison Glover’s classic 1938 study showing local differences in rates of tonsillectomy among British schoolchildren,1 health service researchers have documented extensive variation in the delivery of healthcare in many parts of the world.2 3 4 5 Information on practice variation is important for examining the relations between policy decisions and clinical decisions and raises important questions concerning the efficiency and effectiveness of healthcare. I have therefore argued that population based information should be routinely reported6 and, through the Dartmouth Atlas Project, have taken steps to make such information publicly available for care to Americans aged ≥65 years provided through the Medicare programme. The project has highlighted that much of the variation among hospital referral regions in per capita spending, resource allocation, and service use is unwarranted because it isn’t explained by illness or patient preference.7 8
The publication of The NHS Atlas of Variation in Healthcare,9 which compares healthcare delivery among primary care trusts, shows a similar story in England. Below, I consider the relevance of the Dartmouth work for the NHS.
Unwarranted variation and categories of care
In evaluating practice variation, clinical care can be grouped into three categories with different implications for patients, clinicians, and policy makers:10
Effective care is defined as interventions for which the benefits far outweigh the risks; in this case the “right” rate of treatment is 100% of patients defined by evidence based guidelines to be in need, and unwarranted variation is generally a matter of underuse.
Preference sensitive care is when more than one generally accepted treatment option is available, such as elective surgery; here, the right rate should depend on informed patient choice, but treatment rates can vary extensively because of differences in professional opinion.
Supply sensitive care comprises clinical activities such as doctor visits, diagnostic tests, and hospital admissions, for which the frequency of use relates to the capacity of the local healthcare system. Among older Americans, most of these services are used in caring for chronic illness. However, regions with high rates of use of supply sensitive care do not have better overall outcomes as measured by mortality and indicators of the quality of care, suggesting that the problem in the US is overuse of this category of care.
The Dartmouth atlas and the NHS atlas each report examples of substantial underuse of effective care. Although there is widespread agreement that unvaccinated patients with pneumonia benefit from pneumococcal vaccinations, in some US regions less than 45% of Medicare patients were vaccinated, while in others the figure was over 95%.7 Among primary care trusts, there is more than a fivefold variation in the percentage of diabetic patients receiving nine care processes recommended by the National Institute for Health and Clinical Excellence.9
In the US, underuse of effective care cannot be attributed to lack of resources. Dartmouth atlas studies show that spending more per capita, having more physicians, and producing more hospital admissions (and having a strong national reputation as a high quality academic medical centre) isn’t associated with providing more effective care.11 12 13 So what is associated with this variation in performance? Not surprisingly, the degree to which care is organised and coordinated seems to matter: there is less underuse in regions served by organisations such as the Mayo Clinic or Intermountain Healthcare, where specialists, primary care doctors, and others practise “team medicine.” There is also less underuse in regions where care may be easier to coordinate: those that have fewer doctors managing the care of chronically ill patients and have more primary care doctors relative to the number of specialists.8
Preference sensitive care
Both the US and the English data show extensive variation in elective surgery rates—variation that is much greater than can reasonably be explained by illness. For example, among Americans aged ≥65 years, age, sex, and race adjusted rates of arthritic knee replacement varied by a factor of 4.7, hip replacement by 5.3, and mastectomy for breast cancer by 7.3. The variation among English primary care trusts for knee and hip replacement is lower but is still more than twofold. Spending on hysterectomy shows a fourfold variation among primary care trusts.
It is important to remember that patients in regions with low elective surgery rates are not necessarily going untreated—often, they are being treated differently. Some patients with arthritis of the knee or hip get non-surgical treatments or try lifestyle modifications. Patients with early stage breast cancer are sometimes treated with lumpectomy. Women with abnormal uterine bleeding may have less invasive procedures than hysterectomy. Although patients should choose their treatment, in everyday practice they delegate decision making to physicians, and the decision is therefore strongly influenced by local medical opinion.
Attempts to reduce elective surgery in high rate areas by setting budgets may therefore be unsuccessful as they will not tackle the flaws in clinical decision making that are behind much of the variation. And adherence to evidence-based practice guidelines, as important as these may be, won’t match patients to their preferred treatment. In a study in Ontario, Canada, only 15% of patients who met the clinical guidelines for hip or knee arthroplasty (based on symptom level and radiological changes) actually wanted surgery when asked which treatment they preferred.14 Treating patients according to their preferences—and not giving them treatments they do not want—requires a clinical environment that supports shared decision making and encourages the active engagement of patients in the choice of treatment. How can we achieve this?
Clinical trials show that patient decision aids can improve the quality of clinical decision making, resulting in more knowledgeable patients and treatment decisions more in keeping with the patient’s underlying preferences. Moreover, the use of decision aids seems to decrease the demand for invasive treatments, suggesting that the “right” rate—a rate based on informed patient choice—may be lower than the rate achieved when doctors determine the treatment without the patient’s active participation.15 In an era of tightened budgets and increasing concerns over healthcare rationing, this is good news.
Public support for shared decision making seems to be growing in the UK and the US. It has a key role in the coalition government’s health reform policy16 and is supported in the recent Patient Protection and Affordable Care Act in the US. Broad international support for shared decision making seems also to be growing, as shown at a recent seminar sponsored by the Salzburg Global Seminar.17
Supply sensitive care
Supply sensitive care describes a group of services that are directly related to the supply of physicians, healthcare facilities, and medical equipment. In the US, per capita frequency of use of this kind of care varies strikingly, and most of it is used to manage chronic illness.8 For example, during the last six months of life, the number of physician visits per patient aged ≥65 varied nearly fivefold among regions, from 12.6 to 60.2 visits. The intensity of inpatient care also varied extensively: the proportion of chronically ill patients who had been admitted to an intensive care unit at the time of death ranged from 31.1% to 4.2% (2007 data). In the NHS emergency hospital admissions for elderly people (≥76 years) vary more than twofold, as do admissions for cancer and chronic pulmonary obstructive disease; the percentage of deaths that occur in hospital ranges from 44% to 76%.
There are good reasons for reducing unwarranted variation in supply sensitive care. In the US, it accounts for most of the twofold regional variation in total per capita spending on patients aged ≥65. Understanding why some US regions have more resources (hospital beds, physicians, imaging machines, etc) per capita, and thus spend more on healthcare, requires local knowledge about the dynamics of growth and the incentives inherent in fee for service medicine. The key question, however, is whether more is better; is greater care intensity associated with better health outcomes? Is healthcare being inappropriately rationed in low rate regions or overused in high rate regions?
Unfortunately, patient level clinical research provides little help in answering this question: it is silent on the optimum frequencies of use of supply sensitive care for chronic conditions. Why, then, do I believe that the evidence points to overuse in the US—that more care isn’t better? The evidence comes from studies of outcomes for patients admitted to hospital for hip fractures, heart attacks, and colon cancer.18 Those living in regions with a high intensity pattern of care—who, by virtue of where they are treated, receive more visits, imaging examinations, and admissions—have worse or no better survival than those living in low intensity regions.8 A similar pattern is seen among those who receive their care in academic medical centres.19
While the implications of greater intensity of care are obvious for cost, the evidence for added value in the US is conspicuously absent. Of course, the UK, which has a lower level of resource investment, may not show similar results. But the most important challenge to the clinical and research communities on both sides of the Atlantic is to rationalise the clinical pathways for managing chronic disease: to undertake the clinical research required to convert the “black box” of supply sensitive care into evidence based care that is effective or preference sensitive. This means that the focus of improvement must be on care provided over time (throughout the course of a chronic illness, not just terminal care); it must include all important chronic conditions, not just selected ones, and integrate care among all sectors (community care, inpatient care, extended care, nursing home care, and hospice care). In short, reform requires the integration of primary and specialty care into organised systems capable of coordinating care; rationalising the clinical pathways for managing the population of those with chronic illnesses; and adjusting capacity to reflect the requirements for efficient use of resources.
The advantages to patients as well as the US economy are evident. If the whole of the US followed the practice patterns of high quality, low cost regions served by organised systems such as Mayo Clinic and Intermountain Healthcare, Dartmouth atlas benchmarks suggest it would save 40% of resources spent on chronic illness.20 I suspect that coordination of primary and specialty care would also result in appreciable savings in England.
Putting knowledge into practice
Although research into practice variation has improved our understanding of the nature of unwarranted variations and how they may be reduced, this knowledge has had only modest effects. As the editors of the NHS atlas, Muir Gray and Philip DaSilva, state: “To generate widespread change, the need to identify and reduce unwarranted variation must be placed at the centre of commissioning decision-making, and also needs to be a priority for clinicians and patients.” At a time of shrinking budgets, rising concern over the cost and value of healthcare, and widespread consensus that healthcare must be reformed, motivation to take the painful steps required to change healthcare has never been greater.
The NHS atlas shows how the public sector can ensure the availability of data for monitoring practice variation. In the US and other western nations, access to such data currently depends on the initiatives of researchers, the support of funding agencies, and the willingness of insurers and others to share their data. My hope is that other countries will emulate the NHS example and work together to build an international framework for evaluating heath system performance.
Cite this as: BMJ 2011;342:d1513
Competing interests: The author has completed the unified competing interest form at www.icmje.org/coi_disclosure.pdf (available on request from him) and declares no support from any organisation for the submitted work; JEW is a paid consultant to Health Dialog and the Foundation for Informed Medical Decision; and no other relationships or activities that could appear to have influenced the submitted work.
Provenance and peer review: Commissioned; externally peer reviewed.