45 Validation of a Clinical Decision Rule to Predict Abnormal Thyroid-Stimulating Hormone Levels in Patients Presenting to the Emergency Department With Atrial Fibrillation

Abstract

Guidelines commonly recommend obtaining thyroid-stimulating hormone (TSH) levels in all patients presenting with atrial fibrillation, given that both hyperthyroidism and hypothyroidism have been associated with this arrhythmia. In 2010, Buccelletti et al derived a clinical decision rule that recommends obtaining TSH levels only in patients with any one of the following: previous stroke, hypertension, or thyroid disease. They report a sensitivity of 93% and specificity of 31% for detecting TSH levels <0.35 μIU/mL. We aim to externally validate this clinical decision rule. We conducted a retrospective cohort study of patients who presented to an academic tertiary care emergency department (ED) with atrial fibrillation. Consecutive patients presenting from January 2011 to September 2012 with a final diagnosis of atrial fibrillation were included. Patients were excluded if they did not consent to having their medical records reviewed for research purposes. Charts were reviewed for presence of congestive heart failure (CHF), hypertension, diabetes mellitus, thyroid disease, and cerebrovascular events, as well as age, vital signs, and TSH level. These data were used to assess the sensitivity and specificity of the previously derived clinical decision rule. Additionally, we assessed the sensitivity and specificity of the same model for predicting any abnormal TSH level (TSH <0.3 or >5 μIU/mL). Results are presented as sensitivity, specificity, and positive and negative predictive value (PPV, NPV) with 95% confidence intervals (CI). Of 999 patients who presented to the ED with a final diagnosis of atrial fibrillation, 725 (72.6%) had a TSH level obtained and were included in the analysis. Sixty-eight percent of these levels were drawn in the ED, and the remaining within thirty days of presentation. The overall prevalence of a low TSH was 2.1% (n=15). High TSH levels were identified in 10.9% (n=79). There was no significant association between an abnormal TSH level and age, sex, vital signs, CHF, diabetes mellitus, history of cerebrovascular event, or history of hypertension. History of thyroid disease was significantly associated with an abnormal TSH level (25.5% versus 10.1%, OR 2.7, 95% CI 1.7 to 4.3, P< 0.0001). Applying the rule, 528 of the 725 patients met criteria for having a TSH level drawn, which identified 13 of the 15 patients who had a low TSH. The clinical decision rule had the following diagnostic performance for identifying a low TSH: sensitivity 86.7%, specificity 27.5%, PPV 2.5%, NPV 99.0%. Application of the rule identified sixty of the seventy-nine patients who had a TSH level above the upper limit of normal. When analyzed for its ability to identify high or low TSH levels, the clinical decision rule had the following diagnostic performance: sensitivity 77.7%, specificity 27.9%, PPV 13.8%, NPV 89.3%. The sensitivity of a clinical decision rule including criteria of history of thyroid disease, hypertension, or stroke was found to be 86.7% for identifying low TSH and 77.7% for identifying any TSH abnormality, which was lower than that reported in the original study. A history of thyroid disease was the only historical feature or objective finding significantly associated with a low or abnormal TSH.