Why is there variation?

Almost all of these hospitals are academic medical centers affiliated with medical schools, expected to be the source of medical evidence. Therefore, we would expect best practices to dictate when patients should be admitted to the hospital, how long patients should stay there, and how often they should see their physicians. The degree of variation suggests something else: that physicians are trained to treat patients differently at these various institutions. This variation can be explained in part by the concept of “supply-sensitive care.”

Dartmouth Atlas researchers have defined three categories of care: effective care, preference-sensitive care, and supply-sensitive care. Effective care refers to services that are of proven value and have no significant tradeoffs; that is, the benefits of the services so far outweigh the risks that all patients with specific medical conditions should receive them. Preference-sensitive care is elective care for conditions for which there is more than one treatment option, and patients’ preferences should guide decision-making. Examples include elective surgery and cancer screening tests like mammography and prostate-specific antigen (PSA) testing. The final category is supply-sensitive care, services for which the supply of physicians and resources strongly influences the amount of care delivered. As Figures 3 and 4 show, in areas with more doctors and more hospital beds, patients receive more intense care.

Figure 3.

The relationship between the supply of hospital beds (1996) and hospital discharges per 1,000 Medicare enrollees among hospital referral regions (1995-96).

Figure 4.

The relationship between the supply of cardiologists and visits to cardiologists per 1,000 Medicare enrollees among hospital referral regions (1996).

Figure 3 shows the relationship between hospital beds and admissions for the 306 Dartmouth Atlas hospital referral regions (HRRs), with each blue dot representing a different HRR (please refer to Appendix A for details on how the regions were created). For medical conditions, as the number of hospital beds increases, the admission rate (represented here as the discharge rate) increases. This graph shows that the available supply of hospital beds influences a physicians’ decision to admit a patient.

The red dots at the bottom of the graph show that the rate of hospitalization for hip fracture is not correlated with the supply of hospital beds. This is because admission for hip fracture is effective care; medical evidence shows that it is essential. Therefore, regardless of the availability of beds, patients with hip fractures are hospitalized.

The supply of physicians also influences a patient’s level of care. Figure 4 shows that the number of visits patients make to cardiologists is positively correlated with the number of cardiologists in an HRR. In areas where more cardiologists are practicing, patients are seen more often. This is because in these areas, cardiologists are competing with each other to see a relatively fixed population of patients. In order to keep a full schedule, physicians in these areas will therefore need to see their patients more frequently. This is often not a conscious decision; however, in the absence of clinical evidence surrounding the ideal frequency of physician visits for medical conditions, physicians’ availability influences how often they see their patients for follow-up visits. As suggested in the graph at right, the size of their patient panels affects their judgment, which is in turn affected by the total number of physicians in an area.

Just as the total supply of cardiologists in a given region affects an individual cardiologist’s decision-making, the supply of beds and hospital-based physicians and specialists affects a resident or hospitalist making decisions about the care of an inpatient. The two hospitals in Table 1 ranking highest for most of the variables are in New York City and Los Angeles, cities with high concentrations of hospitals and physicians per capita. The increased supply of resources contributes to the higher utilization in these hospitals.

The high variation among the hospitals in Table 1 shows that we lack the clinical science that could guide practice patterns. Even most of the hospitals at the bottom of this list do not have a best practice strategy regarding the allocation of resources that they could share with higher-intensity hospitals. Physicians working within these hospitals are often unaware of nationwide differences in practice. In fact, when physicians move from a high-intensity hospital to a low-intensity hospital or vice versa, some gradually to adopt the practice patterns prevalent in the new hospital.¹ In other words, the capacity of such resources as hospital beds, along with established practice style, dominates clinical decision-making.

The supply of resources also appears to be more powerful than patient preferences in guiding health care delivery. Patient preferences should strongly influence end-of-life care. Unfortunately, we find that the supply of resources, rather than an individual’s values and desires, dictates the level of care he receives. This was demonstrated by the SUPPORT study, a two-year prospective observational study (phase 1) followed by a two-year controlled clinical trial (phase 2). The first phase indicated that patients preferred less care than they received at the end of their lives. Phase 2, during which patients were randomized to an intervention group and a control group, showed that even after efforts were made to improve communication between the physician, the patient, and the patient’s family about these preferences, patients still received care that they did not desire.² In a follow-up study, Pritchard et al demonstrated that the supply of beds and resources was more powerful in influencing clinical decision-making than patient preferences.³

Unfortunately, more care does not appear to be better, as outcomes remain unchanged, costs increase, and patients report worse experience. Later in the discussion, we explain the research conducted through the Dartmouth Atlas, which indicates that more care may not be better.

Why is this variation relevant to a medical student or resident?

A resident’s training will certainly be affected by the intensity of care provided by the hospital in which he or she trains. The variables discussed in this report capture the ways that physicians in different institutions approach end-of-life care. The variables highlight differences among institutions in the following ways:

Organization of care and reliance on specialists: Complex patients are often cared for by multiple physicians, each having a specific set of recommendations for one aspect of the patient’s medical care. The question is often raised: who is really in charge of this patient’s overall care? The primary team, a primary care physician in the outpatient setting, or a team of residents in the inpatient setting is most frequently charged with the task of consolidating these recommendations and organizing the patient’s care. However, this can be a challenging responsibility. At NYU Medical Center, almost 70% of patients saw ten or more different physicians during their last six months of life. It is probably difficult for both a patient and the primary provider to keep up with the opinions and advice from these ten different physicians. Having so many physicians involved can lead to fragmented or disorganized care. A resident at a hospital where patients have multiple physicians will need to make a concerted effort to keep track of the patient’s care so that the patient does not suffer the adverse consequences of multiple competing, and possibly conflicting, recommendations and treatments.

A high percentage of patients seeing ten or more physicians during the last six months of life also creates a greater reliance on specialists. The benefit of increased specialty care is that a resident at NYU Medical Center will likely interact more frequently with specialists, and she may have the opportunity to witness more procedures than a resident at OHSU, where only 34% of patients saw ten or more physicians in the last six months of life. However, the disadvantage is that her patient’s care will be heavily dictated by specialist opinions. In contrast, a resident at OHSU may be more likely to develop experience managing complex chronic illnesses, as she will be less likely to seek assistance from a specialist.

Utilization of resources at end of life: If asked how they would like to spend their last six months of life, patients are likely to give a wide range of answers. However, the Dartmouth Atlas shows that in many cases, the hospital where a patient receives most of his care determines how he spends these last few months of life. For example, patients at Cedars-Sinai Medical Center saw physicians 82 times in their last six months of life, compared to patients at OHSU, who saw physicians 20 times during that same time period. Similarly, patients at New York-Presbyterian Hospital spent, on average, 21.8 days in the hospital during their last six months of life, compared to 10.6 days at DHMC. These data suggest that the environments at Cedars-Sinai and New York-Presbyterian are more aggressive, leading to more inpatient and outpatient care. More time in the hospital and more physician visits provide more information, allowing a resident to be more certain about a diagnosis and treatment plan or regarding the stability of a patient prior to discharge. However, we know that elderly patients admitted to the hospital are at risk for nosocomial infections and other hazards of hospitalization. In addition, the number of hospital days is highly correlated with inpatient spending. Longer and more frequent hospital stays can lead to worsening illness and increased financial burden for society and for patients who must pay co-insurance and deductibles. In addition, more frequent physician visits can lead to uncoordinated care.

Use of ICUs and reliance on hospice: Many of us were drawn to medicine because we wanted the chance to “save lives,” to engage in heroic measures to cure a disease or stop the bleed, so that a patient can go back to living a great life. Many of these heroic measures happen in an ICU setting, where massive efforts are made to resuscitate patients and keep them alive. There are times, however, when these measures are more harmful than they are heroic—and they may be unwanted by the patient.

The percent of deaths associated with an ICU admission at NYU Medical Center was almost twice what it was at Dartmouth during the period from 2003 to 2007. Days spent in an ICU are very expensive and resource-intensive. Patients are examined and tested multiple times a day, and families often have to drive long distances to see the patient, who may or may not be able to communicate with them. For these reasons, it is an unpleasant experience for the patient and his family. With all patients, but with elderly patients especially, we should ask ourselves if the possible benefits of an ICU stay outweigh these disadvantages. We should be prepared to consider and discuss whether a patient would be better managed at home, perhaps in hospice. In hospitals where there is a high percentage of death in the hospital and in the ICU, there seems to be less of an emphasis on hospice care. Forty-nine percent of patients treated at OHSU were enrolled in hospice in their last six months of life, compared to only 22.5% of patients treated at Cedars-Sinai. A Cedars-Sinai resident may therefore develop a higher threshold for referral of a patient to hospice. He may decide to explore more aggressive treatment approaches before discussing this “comfort care” option with the patient. An OHSU resident may be better trained in having discussions with the patient about his or her preferences for end-of-life care.

The variables in this report show that residents at high-spending hospitals are trained to provide more care to their patients. If increased care leads to better outcomes and happier patients, it might be worth utilizing these resources. However, this is not the case.

Is more care better?

Studies done through the Dartmouth Atlas have shown that increased care improves neither survival nor quality of care. In a cohort study, Dr. Elliott Fisher and his colleagues studied whether patients with similar baseline health status experienced better quality, access, outcomes, and satisfaction in areas of high versus low end-of-life spending. The patient population studied consisted of patients hospitalized between 1993 and 1995 for hip fracture, colorectal cancer or acute myocardial infarction (AMI), along with a sample of the general population from the Medicare Current Beneficiary Survey. The researchers found that patients in higher-spending regions were provided more care in the form of physician visits, hospital days, specialist consultations, and procedures. However, the quality of care, measured by aspirin use for an AMI or rate of influenza immunization, was the same or worse in higher-spending regions compared to lower-spending regions. Similarly, outcomes, measured by the five-year mortality rate, were slightly worse in higher-spending regions in both the colorectal cancer and AMI cohorts. Access and satisfaction with care were found to be the same between high and low-spending regions.^iv, ^4,5 In addition, Medicare spending has been shown to be inversely related to access to specialist care and to patient satisfaction, measured with a global rating of care.^6,7 Finally, quality and satisfaction are positively correlated, meaning that, in those hospitals with low quality measures, patient ratings are also low.⁸

Why do these measures of outcomes, quality, access, and satisfaction worsen as spending and care intensity increase? A plausible hypothesis is that, as spending increases because of higher utilization of resources and the involvement of multiple physicians, care becomes more disorganized. A patient may receive more services than he needs and be exposed to more medical errors, simply because he is getting more care. In addition, without effective coordination of his care, a patient’s preferences may not be heard.