Volume and mortality in coronary artery bypass graftingBMJ 1995; 311 doi: https://doi.org/10.1136/bmj.311.7015.1304b (Published 11 November 1995) Cite this as: BMJ 1995;311:1304
- Harold S Luft,
- Jennifer D Parker
- Professor of health economics Pew postdoctoral fellow Institute for Health Policy Studies, University of California San Francisco, San Francisco, CA 94109, USA
EDITOR,--Amanda J Sowden and colleagues' meta-analysis of six studies of the relation between the volume of coronary bypass graft surgery performed and mortality concludes that the relation is less apparent with better adjustment for case mix and has becomne attenuated over time.1 Unfortunately, in the one study with clinical risk factors, which was also the most recent one and was carried out in New York, 96.7% of all patients were treated in the 26 (out of 30) hospitals with >200 cases. This makes it difficult to detect a volume effect in this study. Without additional studies with clinical risk factor models from other times and places it is impossible to test the authors' hypotheses.
We have addressed the question of a time trend, looking at Californian data for 1983-9. Adjusted death rates by volume of coronary artery bypass grafting were estimated with a risk model based on data from discharge abstracts.2 (Although better risk models may alter these results, they largely reflect substantial differences in mortality rather than in expected rates; predicted risk falls with volume up to about 250 cases a year and is flat thereafter.) Following the method of Sowden and colleagues, we used logistic regression to estimate odds ratios between volume and adjusted risk of death for each year and to determine whether the relation between volume and outcome changed over time. We found significantly lower death rates adjusted for risk in high volume hospitals for each year except 1983 (1983, odds ratio 0.87 (95% confidence interval 0.73 to 1.04); 1984, 0.83 (0.71 to 0.99); 1985, 0.70 (0.60 to 0.81); 1986, 0.77 (0.67 to 0.88); 1987, 0.82 (0.71 to 0.94); 1988, 0.85 (0.74 to 0.96); 1989, 0.85 (0.74 to 0.97)). The interaction of year and volume was not significant. Furthermore, in California the proportions of both hospitals and patients in high volume hospitals were far lower than those in New York and fell during the period, from 54% to 32% and from 78% to 62% respectively.
The figure shows the relation between volume and the ratio of observed to expected mortality for each hospital year of observation. A locally weighted regression scatterplot smoothing curve fitted to the data shows a relatively weak relation at high volumes but a much stronger effect at low volumes, epecially those below 150 cases. These data suggest that concerns about the volume of coronary artery bypass grafting are primarily relevant for the hospitals with very low volumes that are common in places like California but may not exist in other states or countries. Generalisations from meta-analyses that do not take into account the sources of the data may therefore be misleading.
Log in using your username and password
Log in through your institution
Register for a free trial to thebmj.com to receive unlimited access to all content on thebmj.com for 14 days.
Sign up for a free trial