Background: Lack of practical consensus regarding routine electrocardiogram
(ECG) ordering in primary care led us to hypothesize that nonclinical vari
ations in ordering would exist among primary care providers.
Methods: We used 2 computerized billing systems to measure ECG ordering at
visits to providers in 10 internal medicine group practices affiliated with
a large, urban teaching hospital from October 1, 1996, to September 30, 19
97. To focus on screening or routine ECGs, patients with known cardiac dise
ase or suggestive symptoms were excluded, as were providers with fewer than
200 annual patient visits. Included were 69 921 patients making 190 238 vi
sits to 125 primary care providers. Adjusted rates of ECG ordering accounte
d for patient age, sex, and 5 key diagnoses. Logistic regression evaluated
additional predictors of ECG ordering.
Results: Electrocardiograms were ordered in 4.4% of visits to patients with
out reported cardiac disease. Among the 10 group practices, ECG ordering va
ried from 0.5% to 9.6% of visits (adjusted rates, 0.8%-8.6%). Variations be
tween individual providers were even more dramatic: adjusted rates ranged f
rom 0.0% to 24% of visits, with an interquartile range of 1.4% to 4.7% and
a coefficient of variation of 88%. Significant predictors of ECG use were o
lder patient age, male sex, and the presence of clinical comorbidities. Add
itional nonclinical predictors included Medicare as a payment source, older
male providers, and providers who billed for ECG interpretation.
Conclusions: Variations in ECG ordering are not explained by patient charac
teristics. The tremendous nonclinical variations in ECG test ordering sugge
st a need for greater consensus about use of screening ECGs in primary care
.