Digital Breast Tomosynthesis Use, Facility Mammography Volume, and Timely Follow-Up After Abnormal Screening Mammograms in Appalachia, United States.

BACKGROUND

Diagnostic follow-up times after abnormal screening mammography vary widely across facilities. However, whether the use of digital breast tomosynthesis (DBT) is associated with follow-up timeliness, independent of facility context factors, remains unclear.

OBJECTIVE

To evaluate the extent to which facility mammography volume accounts for the association between facility-level DBT use proportion and recall delay after abnormal screening mammography in Appalachia.

METHODS

We linked 2016-2018 Medicare fee-for-service claims to facility survey data from screening facilities in five Appalachian states (KY, OH, PA, VA, and WV). Recall delay was modeled using linear mixed-effects models with random intercepts for facility and beneficiary. The models adjusted for DBT use proportion, capacity score, age, race, dual Medicare-Medicaid eligibility, deprivation, rurality, and stand-alone status. To assess whether facility mammography volume accounted for the association between DBT use proportion and recall delay, we compared two models fitted to the same analytic sample with the same covariates. The models differed only in whether facility mammography volume was included. We estimated 95% confidence intervals for the difference in the DBT coefficient using 5000 bootstrap resamples. Facility-level models and exploratory mediation analyses were conducted as complementary analyses.

RESULTS

DBT use proportion was associated with higher facility mammography volume after adjustment for capacity score and facility covariates (β = 0.614, 95% CI [0.170, 1.058], p = 0.007). After adjustment for capacity score, a 1-SD higher facility mammography volume was associated with shorter recall delay (β = -1.103 days, 95% CI [-2.074, -0.132], p = 0.028). The DBT coefficient changed from -0.970 days before adding facility volume to -0.402 days after volume adjustment, corresponding to a raw coefficient change of 0.568 days (95% bootstrap CI [0.130, 1.088]). The complementary facility-level model estimated a corresponding raw coefficient change of 0.620 days (95% bootstrap CI [0.149, 1.143]). An exploratory mediation analysis adjusted for capacity score and facility covariates estimated an indirect estimate through facility mammography volume of -0.620 days (95% CI [-1.111, -0.146], p = 0.004). The direct and total estimates were not statistically significant.

CONCLUSIONS

Facility-level DBT use proportion was positively associated with facility mammography volume, and higher facility mammography volume was associated with shorter recall delay after adjustment for the capacity score. The association between DBT use proportion and recall delay was not statistically significant either before or after volume adjustment, although the estimated DBT coefficient decreased in magnitude after facility mammography volume was added. These findings suggest that differences in follow-up timeliness may reflect the facility context in which DBT is implemented and support further study of facility-level determinants of diagnostic follow-up.