Correspondence
Volume and Outcome
N Engl J Med 2002; 347:693-696August 29, 2002
- Article
To the Editor:
The attempt of Birkmeyer et al. (April 11 issue)1 to correlate low procedure-specific hospital volume with increased mortality has methodologic and interpretive problems. The investigators used data from the Medicare Provider Analysis and Review (excluding those for Medicare patients enrolled in health maintenance organizations) and from the Nationwide Inpatient Sample, without verification, to estimate procedure-specific hospital volume. The correlation between Medicare volume and hospital volume (correlation coefficient, 0.97) probably resulted from mathematic coupling,2,3 which occurs when variables are shared. Were the relations between outcome and volume significant when only Medicare volumes were analyzed?
No proof of validation of the regression models is presented. For most procedures, lower-volume institutions had higher percentages of nonelective admissions, patients over 75 years of age, and black patients (a fact that the authors erroneously interpret as indicating that black patients were more likely to be treated at low-volume hospitals). These three variables are included in the regression analysis, but they are the tip of the iceberg made up of a multitude of other unreported confounders — such as preoperative selection of patients, intraoperative management, and postoperative care. Obviously, it is impossible to assign patients randomly to hospitals, but without robust, validated regression equations, the relative importance of volume may be overestimated or underestimated.
3 ReferencesJames E. Barone, M.D.
Stamford Hospital, Stamford, CT 06902
drjbarone@stamhosp.chime. org Donald A. Risucci, Ph.D.
John A. Savino, M.D.
New York Medical College, Valhalla, NY 105951
Birkmeyer JD ,Siewers AE ,Finlayson EVA , et al. Hospital volume and surgical mortality in the United States. N Engl J Med 2002;346:1128-1137
Full Text | Web of Science | Medline2
Archie JP Jr . Mathematic coupling of data: a common source of error. Ann Surg 1981;193:296-303
CrossRef | Web of Science | Medline3
Barone JE ,Lowenfels AB . Maximization of oxygen delivery: a plea for moderation. J Trauma 1992;33:651-653
Web of Science | Medline
To the Editor:
The article by Birkmeyer et al. advances our understanding of the relation between hospital volume and outcome. We are curious, however, about whether the authors attempted to analyze surgical mortality according to the patient's level of surgical risk. In the case of coronary-artery bypass grafting, we have shown that differences in mortality rates between low-volume hospitals and high-volume hospitals might be driven predominantly by differences among patients at high surgical risk.1 Patients who are at low risk, in contrast, might receive little or no benefit from obtaining their care at high-volume centers. This possibility has clear implications for regionalization policies, in general, and for high-volume procedures such as coronary-artery bypass grafting, in particular.2 To avert an estimated 314 annual deaths related to coronary-artery bypass grafting, for example, very-high-volume centers would need to double their capacity by absorbing more than 31,000 additional cases per year. If high-risk patients who would clearly benefit from the expertise available at high-volume centers could be identified preoperatively and then selectively sent to such regional centers, these logistic problems would be greatly diminished.
Finally, we would caution against extrapolation of these data to patients younger than 65 years of age. Advanced age — both directly and in association with coexisting conditions — clearly increases a patient's base-line risk of death related to surgery. The population studied by Birkmeyer et al. therefore represents a high-risk group of patients, as reflected in the relatively high 30-day mortality reported for patients undergoing coronary-artery bypass grafting: 4.8 percent at very-high-volume centers, as compared with a national average of 2.9 percent.3
3 ReferencesBrahmajee K. Nallamothu, M.D., M.P.H.
University of Michigan Medical Center, Ann Arbor, MI 48109-0022
bnallamo@umich.edu Sanjay Saint, M.D., M.P.H.
Ann Arbor Veterans Affairs Medical Center, Ann Arbor, MI 48105Kim A. Eagle, M.D.
University of Michigan Medical Center, Ann Arbor, MI 48109-00221
Nallamothu BK ,Saint S ,Ramsey SD ,Hofer TP ,Vijan S ,Eagle KA . The role of hospital volume in coronary artery bypass grafting: is more always better? J Am Coll Cardiol 2001;38:1923-1930
CrossRef | Web of Science | Medline2
Luft HS . Better for whom? Policy implications of acting on the relation between volume and outcome in coronary artery bypass grafting. J Am Coll Cardiol 2001;38:1931-1933
CrossRef | Web of Science | Medline3
Grover FL . The Society of Thoracic Surgeons National Database: current status and future directions. Ann Thorac Surg 1999;68:367-373
CrossRef | Web of Science | Medline
To the Editor:
Aside from the direct costs of medical care, travel to a high-volume center can be costly and is not affordable for all patients. An uninsured patient with pancreatic cancer will probably not be accepted at a high-volume institution and will most likely not even be offered the option of going to one. Those who can afford the expense of copying medical, pathological, and radiologic records, and traveling for the consultation, procedure, and any follow-up will do so, leaving those who are less fortunate at the local center. Birkmeyer et al. controlled for coexisting conditions, but I would bet that patients who are able to travel are patients with a better prognosis.
Christopher K. Senkowski, M.D.
Mercer University School of Medicine, Savannah, GA 31404
senkoch1@memorialhealth.com To the Editor:
The studies by Birkmeyer et al. and Begg et al. (April 11 issue)1 add further evidence of the correlation between volume and quality in the delivery of certain health care services. Whereas others will certainly challenge the validity or reliability of the data sets or will question whether adequate risk adjustment would temper or invalidate the results, I instead question the policy suggestions laid out in the accompanying editorial by Epstein.2 It seems much too premature to recommend broadly the diversion of selected procedures from low-volume hospitals, even as a “transitional strategy.” Certainly, experience matters, but what else is included in the equation? It is still unclear whether volume itself is a generalizable predictor of quality at all, or whether and how much such underlying factors as organizational design, streamlined data management, or multidisciplinary care act as confounders. Without such knowledge, merely stripping further volume away from low-volume centers may be strikingly counterproductive in terms of the goal of high quality overall.
2 ReferencesDarren M. Kocs, M.D.
University of Michigan Medical Center, Ann Arbor, MI 48109
dkocs@umich.edu 1
Begg CB ,Riedel ER ,Bach PB , et al. Variations in morbidity after radical prostatectomy. N Engl J Med 2002;346:1138-1144
Full Text | Web of Science | Medline2
Epstein AM . Volume and outcome -- it is time to move ahead. N Engl J Med 2002;346:1161-1164
Full Text | Web of Science | Medline
To the Editor:
Although we generally agree with Epstein's support of targeted policies to decrease the proportion of surgical procedures performed at low-volume hospitals, such policies trouble us because volume seems a fairly crude predictor of patient outcome and the closely related measure of the quality of health care. Why not try to identify the true determinants of patient outcomes or quality of health care? Such information could be used to predict the outcome for patients treated by a particular type of physician at a particular type of hospital, reveal characteristics of the physicians and hospitals from which the best outcomes could be expected, and design interventions to improve quality (for all types of health care workers and health care facilities — not just surgeons working in hospitals).
Alexander K. Rowe, M.D., M.P.H.
Michael S. Deming, M.D., M.P.H.
Centers for Disease Control and Prevention, Atlanta, GA 30341
axr9@cdc.gov To the Editor:
Epstein raises some of the key dilemmas facing policy makers as they wrestle with the implications of the association between volume and outcome. In particular, interventions directed at influencing the referral of patients toward high-volume institutions, by fiat, incentive, or consumer pressure, leave unanswered the question of what to do with the lower-volume centers. It is helpful to recognize that volume alone does not presage outcome. The amount of improvement in performance that can be extracted from a given number of procedures also depends in part on how the activities related to learning are managed at a given institution.1 We have found notable variation in the quality and quantity of management attention paid to learning in the case of the adoption of a new form of medical technology.2 Perhaps part of the solution to the policy makers' dilemma is better management.
2 ReferencesRichard Bohmer, M.B., Ch.B., M.P.H.
Amy Edmondson, Ph.D.
Gary Pisano, Ph.D.
Harvard Business School, Boston, MA 021631
Pisano GP ,Bohmer RMJ ,Edmondson AC . Organizational differences in rates of learning: evidence from the adoption of minimally invasive cardiac surgery. Management Sci 2001;47:752-768
CrossRef | Web of Science2
Edmondson AC ,Bohmer RM ,Pisano GP . Disrupted routines: team learning and new technology implementation in hospitals. Administrative Sci Q 2001;46:685-716
CrossRef | Web of Science
To the Editor:
The editorial by Dr. Epstein is laudable in its review of recent information concerning the relation between surgical volume and outcomes, but important points are not well understood. As a case in point, six years ago, I began referring surgical and general medical patients from our small rural hospital to regional referral centers. As this practice led to what I had perceived to be improved patient care, it also created much larger problems in the community: as surgical volume decreased, the skills of the surgical team also deteriorated. The community was then left without a general surgeon. The hospital decreased in size, and the future of this community resource appears to be in danger.
John L. Ghertner, M.D.
6055 Robinson Rd., Sodus, NY 14551
nghertner@earthlink.net To the Editor:
I think the assumption made by Epstein that all low-volume hospitals have high mortality after high-risk surgical procedures is unjustified by the data presented. We in the medical profession have no business getting into this matter. Instead, let us use the data to look at the various hospitals and improve the quality of care. Preventing low-volume hospitals from performing surgery can have dangerous consequences. For instance, preventing low-volume surgeons from performing elective repair of abdominal aortic aneurysms will have negative consequences in the form of patients with ruptured aneurysms. Surgeons who are not performing elective aneurysm repair may not feel comfortable repairing ruptured aneurysms — a procedure requiring far more experience and skill. Patients with ruptured aneurysms may have to travel substantial distances to get to centers where the procedure is performed.
William W. Babson, Jr., M.D.
Jordan Hospital, Plymouth, MA 02360
dkbabson@aol.com Author/Editor Response
The authors reply:
To the Editor: In our study, hospital volume was measured directly by counting Medicare procedures. To facilitate interpretation of our results, we converted hospital Medicare volumes to total hospital volumes, using data from the Nationwide Inpatient Sample to estimate the appropriate multiplier for each procedure. Because the same multiplier was applied to all hospitals, this approach does not in any way affect the magnitude of reported volume–outcome relations. Barone et al. also note the importance of accounting for potentially confounding variables, including variables related to patient selection. As we acknowledged, the extent to which data on claims capture these variables is limited. However, it would be inappropriate to adjust for variables related to operative and postoperative care. Such processes of care are most likely essential parts of the causal pathway underlying the relations between volume and outcome and thus should not be viewed as “confounders.”
Nallamothu et al. wonder whether hospital volume may be most important for high-risk patients, as their research has suggested with regard to coronary-artery bypass grafting.1 Our (unpublished) analyses support this premise when the importance of volume is expressed in terms of absolute differences in operative mortality. Differences in mortality appear to be largest for subgroups of patients with the highest base-line risk, such as the very elderly. In terms of the relative risk of death, however, we found the effect of hospital volume to be relatively uniform among subgroups defined according to base-line risk.
Senkowski suggests that lower-income patients and those without insurance may be less likely to undergo surgery at high-volume hospitals. Although previous studies have demonstrated only moderate differences between high-volume hospitals and low-volume hospitals in the socioeconomic status of patients, we agree that more work is needed to clarify the role of patient selection in the observed relations between volume and outcome. We also agree that policy makers must be careful to ensure that less fortunate patients are not left behind as volume-based referral initiatives are implemented.
Given the hundreds of articles published over the past several decades demonstrating better outcomes with selected procedures at high-volume hospitals,2,3 we do not agree with Kocs that ongoing efforts to translate this information into policy are “premature.” As he points out, hospital volume may serve as a proxy for numerous organizational characteristics associated with better quality. Efforts to identify these attributes and to ensure that they are implemented broadly would be worthwhile. Meanwhile, many unnecessary deaths could be averted by policies concentrating pancreatic resections, esophagectomies, and other high-risk procedures in high-volume centers.
3 ReferencesJohn D. Birkmeyer, M.D.
Dartmouth–Hitchcock Medical Center, Lebanon, NH 03756
john.d. birkmeyer@hitchcock. org Emily V.A. Finlayson, M.D.
University of California at San Francisco, San Francisco, CA 941431
Nallamothu BK ,Saint S ,Ramsey SD ,Hofer TP ,Vijan S ,Eagle KA . The role of hospital volume in coronary artery bypass grafting: is more always better? J Am Coll Cardiol 2001;38:1923-1930
CrossRef | Web of Science | Medline2
Dudley RA ,Johansen KL ,Brand R ,Rennie DJ ,Milstein A . Selective referral to high-volume hospitals: estimating potentially avoidable deaths. JAMA 2000;283:1159-1166
CrossRef | Web of Science | Medline3
Halm EA, Lee C, Chassin MR. Is volume related to quality in health care? A systematic review and methodologic critique of the research literature. Ann Intern Med (in press).
Author/Editor Response
The editorialist replies:
To the Editor: Drs. Babson, Ghertner, and Kocs all express concern about the potentially deleterious consequences of diverting patients from low-volume hospitals. I share their concern. However, the relations between volume and outcome are extraordinarily strong for some surgical procedures. For example, according to data from Birkmeyer et al., risk-adjusted mortality rates for esophagectomy varied from 8.4 percent in very-high-volume hospitals to 20.3 percent in very-low-volume hospitals. Risk-adjusted mortality rates for pancreatic resection and repair of an abdominal aortic aneurysm varied from 3.8 percent to 16.3 percent and from 3.9 percent to 6.5 percent, respectively. Clearly, we need research to understand the causes of these differences and quality-improvement interventions to improve our care and reduce these differences. But the benefit from these activities may not be seen for years. For patients receiving care now, the effect of these differences on patients' health outcomes is too large to ignore.
Policy changes can often result in unanticipated consequences. Therefore, our approach should be demonstration and evaluation with close follow-up and readjustment. I recognize Dr. Ghertner's point that rural hospitals are especially vulnerable and their loss would threaten patients' access to care. I therefore proposed that efforts to divert patients to low-volume hospitals be limited to urban areas, as well as to procedures for which the relations between volume and mortality are strongest, and to institutions at the bottom end of the spectrum with very low volume. I urge that the Leapfrog Group and others who support more aggressive policies refine their current plans as they move toward interventions that go beyond education. Broad-scale regionalization is not warranted at this time.
Dr. Bohmer and colleagues and Drs. Rowe and Deming underscore the fact that volume is merely a characteristic that is associated with differences in patient outcome rather than a direct measure of quality of care. Differences in risk-adjusted mortality arise largely from differences in clinical management. Efforts to improve clinical management, reduce mortality, and diminish differences in the quality of care should be our highest priority.
Arnold M. Epstein, M.D.
Harvard School of Public Health, Boston, MA 02115- Citing Articles (4)
Citing Articles
1
Craig Evan Pollack, Justin E. Bekelman, Andrew J. Epstein, Kaijun Liao, Yu-Ning Wong, Katrina Armstrong. (2011) Racial Disparities in Changing to a High-volume Urologist Among Men With Localized Prostate Cancer. Medical Care1
CrossRef2
Elisabeth T. Tracy, Kyla M. Bennett, Emeline M. Aviki, Theodore N. Pappas, Bradley H. Collins, Janet E. Tuttle-Newhall, Carlos E. Marroquin, Paul C. Kuo, John E. Scarborough. (2009) Temporal trends in liver transplant centre volume in the USA. HPB 11:5, 414-421
CrossRef3
John Varkarakis, Walter Wirtenberger, Germar-Michael Pinggera, Andreas Berger, Toru Harabayashi, Georg Bartsch, Wolfgang Horninger. (2004) Evaluation of urinary extravasation and results after continence-preserving radical retropubic prostatectomy. BJU International 94:7, 991-995
CrossRef4
Tofy Mussivand, Delphine A. Hasle, Kevin S. Holmes. (2004) Is Center Specific Implantation Volume a Predictor of Clinical Outcomes with Mechanical Circulatory Support?. ASAIO Journal 50:1, 33-36
CrossRef
