Table 2.
Mixed-effects modeling of sensitivity, specificity, and positive predictive value 1 by facility characteristics with adjustment for patient and radiologist characteristics and mammography registry (N = 43 facilities with 360 149 screening mammography examinations)*
| Odds of having a negative mammogram given no cancer diagnosis (specificity) | Odds of having a positive mammogram given a cancer diagnosis (sensitivity) | Odds of having a cancer diagnosis given a positive mammogram (PPV1) | ||||
| Facility characteristic | OR (95% CI) | Overall P | OR (95% CI) | Overall P | OR (95% CI) | Overall P |
| Facility structure and organization | ||||||
| Facility volume (average no. of mammograms per year)† | ||||||
| ≤1500 | 1.00 (referent) | .002 | 1.00 (referent) | .097 | 1.00 (referent) | .202 |
| 1501–2500 | 0.65 (0.48 to 0.88) | 2.77 (1.15 to 6.65) | 1.04 (0.70 to 1.55) | |||
| 2501–6000 | 0.66 (0.49 to 0.89) | 2.29 (1.05 to 5.03) | 0.84 (0.58 to 1.21) | |||
| >6000 | 0.53 (0.38 to 0.74) | 2.53 (1.17 to 5.44) | 0.99 (0.68 to 1.43) | |||
| Is your mammography facility for-profit or not-for-profit? | ||||||
| Not-for-profit | 1.00 (referent) | .315 | 1.00 (referent) | .324 | 1.00 (referent) | .057 |
| For-profit | 0.88 (0.67 to 1.14) | 1.26 (0.79 to 2.01) | 0.82 (0.66 to 1.01) | |||
| Does this facility offer diagnostic mammograms? | ||||||
| No | 1.00 (referent) | .003 | 1.00 (referent) | .883 | 1.00 (referent) | <.001 |
| Yes | 0.66 (0.50 to 0.86) | 0.95 (0.50 to 1.83) | 0.63 (0.49 to 0.82) | |||
| Interpretive and audit process | ||||||
| Are any screening mammograms performed at your facility interpreted by a radiologist who specializes in breast care? | ||||||
| No | 1.00 (referent) | .083 | 1.00 (referent) | .652 | 1.00 (referent) | .039 |
| Yes | 1.26 (0.97 to 1.63) | 0.90 (0.57 to 1.42) | 1.23 (1.01 to 1.50) | |||
| What percentage of the screening mammograms done at your facility are interpreted at another facility?‡ | ||||||
| 0 | 1.00 (referent) | .002 | 1.00 (referent) | .093 | 1.00 (referent) | .072 |
| 80–100 | 0.70 (0.56 to 0.87) | 1.59 (0.92 to 2.72) | 0.82 (0.66 to 1.02) | |||
| How are decisions made for mammograms interpreted by more than one radiologist? | ||||||
| Double reads not performed | 1.00 (referent) | .177 | 1.00 (referent) | .779 | 1.00 (referent) | .005 |
| Independent double reads | 0.84 (0.68 to 1.03) | 0.90 (0.63 to 1.30) | 0.90 (0.78 to 1.03) | |||
| Double reads by consensus | 0.76 (0.49 to 1.16) | 1.03 (0.53 to 2.01) | 0.62 (0.48 to 0.82) | |||
| How often is individual radiologist-level audit data given back to radiologists on their performance? | ||||||
| Once a year | 1.00 (referent) | <.001 | 1.00 (referent) | .291 | 1.00 (referent) | .050 |
| Two or more times per year | 1.54 (1.23 to 1.94) | 0.74 (0.50 to 1.09) | 1.23 (1.03 to 1.47) | |||
| Unknown | 0.81 (0.61 to 1.06) | 1.08 (0.60 to 1.96) | 1.13 (0.91 to 1.40) | |||
| How is this audit information reviewed? | ||||||
| Reviewed together (with other radiologists) in meeting | 1.00 (referent) | .001 | 1.00 (referent) | .609 | 1.00 (referent) | <.001 |
| Reviewed by facility or department manager or by lead radiologist alone | 0.60 (0.44 to 0.84) | 1.22 (0.61 to 2.44) | 0.94 (0.69 to 1.27) | |||
| Reviewed by each radiologist alone | 1.32 (0.94 to 1.83) | 1.20 (0.49 to 2.93) | 1.07 (0.73 to 1.58) | |||
| Unknown | 0.70 (0.53 to 0.92) | 1.38 (0.83 to 2.30) | 0.65 (0.54 to 0.79) | |||
*
OR = odds ratio; CI = confidence interval; PPV1 = positive predictive value 1.
†
Based on registry data.
‡
No facilities reported that 1%–79% of the screening mammograms done at the facility were interpreted at another facility.