TABLE 2—
Relationship of Hospital Volume to All-Cause and Breast Cancer–Specific Mortality: United States, 1994–1996
| No. | Deaths From All Causes, % | All-Cause Mortality HR (95% CI) | Breast Cancer Deaths, % | Breast Cancer–Specific Mortality HR (95% CI) | |
| Hospital volumea | |||||
| 0–19 | 3596 | 28.8 | 1.00 | 7.3 | 1.00 |
| 20–39 | 3698 | 25.6 | . . .b | 5.5 | 0.80 (0.66, 0.97) |
| ≥ 40 | 3931 | 23.4 | 0.83 (0.75, 0.92) | 6.2 | 0.78 (0.64, 0.96) |
| Comorbidity index | |||||
| 0 | 8369 | 21.0 | 1.00 | 5.8 | 1.00 |
| 1 | 1983 | 33.4 | 1.60 (1.46, 1.75) | 7.2 | 1.25 (1.03, 1.51) |
| ≥ 2 | 873 | 55.2 | 3.40 (2.71, 3.40) | 9.0 | 1.67 (1.28, 2.17) |
| Tumor size, mm | |||||
| 0–20 | 7790 | 20.2 | 1.00 | 3.5 | 1.00 |
| 21–50 | 3183 | 38.7 | 1.61 (1.48, 1.74) | 12.6 | 2.41 (2.05, 2.84) |
| >50 | 150 | 38.7 | 1.66 (1.27, 2.16) | 16.0 | 4.25 (2.77, 6.53) |
| Missing | 102 | 33.3 | 1.22 (0.86, 1.73) | 11.8 | 1.19 (0.66, 2.15) |
| Lymph nodes | |||||
| 0 | 7897 | 21.1 | 1.00 | 3.8 | 1.00 |
| ≥ 1 | 2293 | 33.7 | 1.59 (1.44, 1.74) | 14.2 | 3.07 (2.59, 3.65) |
| Missing | 1035 | 44.3 | 1.55 (1.37, 1.75) | 11.4 | 1.77 (1.33, 2.37) |
| Estrogen receptor/progesterone receptor status | |||||
| Negative | 1344 | 33.6 | 1.00 | 14.7 | 1.00 |
| Positive | 7741 | 24.0 | 0.66 (0.59, 0.73) | 4.9 | 0.42 (0.35, 0.50) |
| Not available | 2140 | 27.8 | 0.72 (0.63, 0.84) | 6.2 | 0.48 (0.37, 0.62) |
| Tumor grade | |||||
| 1 or 2 | 6077 | 21.9 | 1.00 | 3.7 | 1.00 |
| 3 or 4 | 3016 | 33.6 | 1.38 (1.27, 1.51) | 12.3 | 2.26 (1.89, 2.70) |
| Not available | 2132 | 26.0 | 1.08 (0.97, 1.19) | 5.4 | 1.26 (1.00, 1.60) |
Note. HR = hazard ratio; CI = confidence interval. Values were adjusted for all of the factors shown, as well as for age, race, per capita income, population density, propensity group, and Surveillance, Epidemiology, and End Results program site.
aHospital volume was defined as number of breast cancer operations performed per year.
b Treated as a separate stratum to retain proportionality of hazards in other volume groups and other variables in the model.