Statistical analysis
We examined the association between surgeon-level EMR and surgeon years
in practice via linear regression models. To account for differences in
the model specifications used for California and New York multivariable
risk models, EMRs were standardized by dividing surgeon EMR by the
corresponding state’s average mortality rates for isolated CABG, 1.67%
for New York and 2.44% for California. This yielded standardized risk
relative to each state’s average observed mortality rate. A sensitivity
analysis was also performed to ensure EMR was not skewed by surgeons
with low case volumes. Observed to expected (O:E) ratios were calculated
for each surgeon and analyzed via linear regression with years in
practice, with adjustment for case volume. For this O:E ratio
regression, case volumes were divided by the average CABG case volumes
for the New York and California study periods to account for the longer
study period captured in the New York data. We defined statistical
significance as p <0.05. All data analysis was performed
in R version 3.6.3 (R Foundation for Statistical Computing).