Background: Fine needle aspiration cytology (FNAC) cannot differentiate between benign and malignant conditions in cytologically indeterminate thyroid lesions. Therefore, a minimum of diagnostic lobectomy is required for definitive diagnosis. The objective of this study is to identify the rate of malignancy and clinical features that may possibly predict malignancy in patients with these lesions, in Ethiopian hospitals. Methods: This was a retrospective review of the medical records of patients who underwent surgery for cytologically indeterminate thyroid lesions in three referral hospitals between September 2015 and September 2020. Results: Of 85 patients with indeterminate cytology findings, 56 (63.5%) were follicular, and 29 (34.1%) were reported to be hurthle cell neoplasms. Follicular lesions of undetermined significance (FLUS) and suspicious for follicular neoplasm were each reported in single cases (1.7%). Malignant disease was diagnosed in 19 (22.4%) of patients. A follicular variant of papillary cancer was detected in 7 (11.5%) patients. Hard nodule consistency was reported in 9 of 11 malignant lesions and 5 of 66 benign lesions. In multivariate binary logistic regression, hard nodule consistency was found to be associated with malignancy (P = 0.012, AOR = 7.28 (1.5, 34.54) 95% CI ). The ill-defined surface of a nodule was found to be associated with malignancy though the association was not statistically significant (P = 0.088, AOR = 0.162 (0.020, 1.313) 95% CI. Ultrasound evaluation of thyroid nodule was performed only in 41 (47.7%) of patients. Conclusion: The rate of malignancy in thyroid nodules with indeterminate cytology was 22.4%. The risk of malignancy was higher in patients with hard thyroid nodule consistency and ill-defined surface. Despite the established benefits of ultrasound for the evaluation of thyroid nodules, the current practice of its use in our setup is suboptimal. Keywords: Follicular; hurthle cell; indeterminate cytology; predictors of malignancy