The major finding of this study is the very rapid growth in care by hospitalists in the late 20th and early 21st century. The odds that a hospitalized Medicare patient would receive care from a hospitalist increased by 29.2% per year from 1997 to 2006. The map of hospitalist penetration according to hospital-referral region () shows that by 2006, almost all areas of the United States were served by hospitalists. Also, although the growth of care by hospitalists has been greater in large hospitals, teaching hospitals, and in certain geographic areas, substantial growth occurred in every area and type of hospital.
The growth in care by hospitalists who are general internists resulted in an increase in the percentage of all Medicare inpatients who were cared for by a general internist, whether hospitalist or nonhospitalist, from 46.4% in 1995 to 61.0% in 2006. The increase in care by hospitalists () started before prospective trials and observational studies showed that hospitalists were associated with decreased hospital costs, with no deleterious effect on outcomes.12–17
Saint and colleagues18
made a similar point in their analysis of 1994 Medicare data from Washington State.
Our data also suggest that the career stability of hospitalists has increased over time. Only 33.4% of physicians identified as being hospitalists in 1995 still met the definition 2 years later, whereas 65.7% of hospitalists identified in 2004 still derived 90% or more of their claims from the care of hospitalized patients 2 years later.
This research has limitations. Analyses of Medicare claims will not identify pediatric hospitalists19,20
and hospitalists who only treat patients in health maintenance organizations (HMOs). Our definition also excludes medical subspecialists and family physicians who are hospitalists. A 2006–2007 survey by the Society of Hospital Medicine showed that 82.3% of hospitalists were general internists, 4.0% were internal-medicine subspecialists, 3.7% were family practitioners, and 10% were pediatricians or pediatricians who were also internists.4
Because we were assessing only hospitalists in general internal medicine, our estimates of the percentage of patients cared for by hospitalists were limited to those patients who received care from any physician in general internal medicine while hospitalized. Including all hospitalized patients in the denominator would result in lower estimates of hospitalist care, but it would not substantially alter the association of patient and hospital characteristics with hospitalist use. The only noteworthy change was a much lower estimate for cardiology surgical DRGs.
We considered a physician to be a “general internist” based on the specialty code shown on the majority of his or her Part B claims in a given year. This method has been used in previous studies,21–23
and it may differ somewhat from other sources of specialty information, such as the Physician Masterfile of the American Medical Association and the Unique Physician Identification Number Directory.24
We used a functional definition of hospitalist, whereas previous studies of hospitalists have been based on self-report (i.e., a hospitalist is a physician who calls himself or herself a hospitalist). We chose a cutoff point of 90% for evaluation-and-management claims generated from the care of hospitalized patients because that was the category that grew rapidly from 2001 through 2006. Our estimate of the number of hospitalists nationally was dependent on cutoff points chosen for the algorithms. For example, the estimated number of hospitalists in 2006 was 13,466 with the use of the cutoff point of 90% for evaluation-and-management charges generated from the care of hospitalized patients. At a cutoff point of 50%, the estimate would be 23,112; at 70%, it would be 17,568; and at 80%, it would be 15,560. Many early descriptions of hospitalists were of physicians who devoted more than 50% of their clinical time to inpatient care but continued to provide outpatient care.12,13,18
Our definition would have missed many of these persons. It is not clear to what extent this model is still followed.25
A 2007–2008 survey of hospitalist physicians showed that, on average, only 1.1% of their total encounters were with outpatients and only 0.9% were with patients in the emergency room, suggesting that the great majority of hospitalists today are devoting their time almost exclusively to the care of hospitalized patients.4
The estimate of hospitalists was also dependent on the cutoff point chosen for the minimum number of Medicare evaluation-and-management charges in the 5% Medicare sample in order for a physician to be included in the analyses. If we had changed the cutoff point from 5 evaluation-and-management charges to 10, then the estimated number of hospitalists in 2006 would have decreased from 13,466 to 11,897. However, the use of different cutoff points for the number of evaluation-and-management charges or the percentages of these charges generated from services provided to hospitalized patients did not greatly affect the dramatic growth of hospitalists over the period from 1995 through 2006.
The specificity of our definition seems to have improved over time. As noted in the Methods section, 14.9% of physicians who were identified as hospitalists in 1995 billed for procedures normally performed by subspecialists, and this percentage decreased to 2.3% in 2006.
Another limitation of our study is that some of our inferences are based on indirect data. For example, we inferred from the data in that an absence of Medicare billing indicated that the physician was in training, but retired physicians or full-time employees of HMOs or government hospitals also might not generate Medicare charges.
The rapid growth of care by hospitalists exemplifies the dynamic nature of medical care in the United States in recent years. The decreased variation in rates of hospitalist care over time according to most patient and hospital characteristics is consistent with previous studies of the diffusion of new medical practices and techniques.26,27
Medicare data allow us to track many of those changes at a national level.